WO2017037768A1 - Evaluation system, evaluation method, and data analysis system - Google Patents

Evaluation system, evaluation method, and data analysis system Download PDF

Info

Publication number
WO2017037768A1
WO2017037768A1 PCT/JP2015/074319 JP2015074319W WO2017037768A1 WO 2017037768 A1 WO2017037768 A1 WO 2017037768A1 JP 2015074319 W JP2015074319 W JP 2015074319W WO 2017037768 A1 WO2017037768 A1 WO 2017037768A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
variable
objective variable
unit
period
Prior art date
Application number
PCT/JP2015/074319
Other languages
French (fr)
Japanese (ja)
Inventor
森脇　紀彦
知明 秋富
淳一 平山
矢野　和男
竜治 嶺
文也 工藤
篤志 宮本
Original Assignee
株式会社日立製作所
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 株式会社日立製作所 filed Critical 株式会社日立製作所
Priority to PCT/JP2015/074319 priority Critical patent/WO2017037768A1/en
Priority to JP2016574300A priority patent/JP6326510B2/en
Publication of WO2017037768A1 publication Critical patent/WO2017037768A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services

Definitions

  • the present invention relates to an evaluation system, an evaluation method, and a data analysis system.
  • Patent Document 1 creates a process for sequentially listing explanatory variables that are statistically highly correlated with a set objective variable, and a regression equation for the relationship between the explanatory variable and the objective variable.
  • a data analysis system that performs processing is described.
  • the service effect calculation process for reducing the risk of unknown measure introduction is performed. This process quantitatively evaluates how changes in the behavior of customers and employees due to the introduction of measures contributed to profits.
  • Patent Document 2 refers to the operational impact that reflects how much the operation evaluation index changes due to the change in the value of the operation variable based on the difference between the maximum value and the minimum value of the operation evaluation index for a certain operation variable. I'm looking for something. By extracting an operation variable having a large operation influence degree, an operation variable effective for improving the operation state is selected.
  • Patent Document 3 divides the data relating to the objective variable, calculates the unity degree obtained from the sum of squares of deviation for each of the two divided sets of objective variables, and when the value of the unity degree is large, It is judged that there is a large statistically significant difference between data belonging to the set.
  • Patent Document 1 in order to evaluate the effect of introduction of a measure, a behavior change measure is implemented, and a service effect calculation process is performed after a certain period of time.
  • a behavior change measure is implemented, and a service effect calculation process is performed after a certain period of time.
  • the operation variable is obtained from the maximum value and the minimum value of the operation evaluation index calculated based on the probability density of the determination result of the operation status, and the operation variable effective for improving the operation status is obtained. Seeking.
  • this calculation method it is possible to determine the influence of the operation variable on the operation evaluation index, but it is difficult to evaluate whether there is room for improvement in the operation state itself. For example, it is possible to extract the operating variables that have the greatest effect on the rate of wind reduction, but whether the data analysis system can reduce the rate of wind reduction and maintain good furnace conditions by implementing the measures presented by the data analysis system. Direct judgment was difficult. For this reason, it has been difficult to determine whether the objective variable (operation status) itself has a possibility of improving with the technique described in Patent Document 2.
  • Patent Document 3 a sum of squared deviations is obtained for each set of divided objective variables, and if the sum of the squared deviations of each set is small, statistics between two divided sets of objective variables are obtained.
  • the data analysis is performed based on the two sets generated by the division. That is, as the value of the sum of squared deviations of the data in each set becomes smaller, it is determined that the division points that generate such a set are more suitable for analysis.
  • this method it is possible to appropriately determine the dividing point for analyzing the data of the objective variable, but it is difficult to determine whether there is room for improvement in the objective variable itself.
  • this application evaluates whether the measures proposed by the data analysis system have an improvement effect on the objective variable that is an indicator of the performance specified by the customer before introducing the data analysis system. It is an object of the present invention to provide a technique that makes it easier to display information such as an objective variable and an explanatory variable having a high correlation with the objective variable during a period of high evaluation.
  • An example of a representative means for solving the above problem is an evaluation system for displaying the effect of introducing a data analysis system, and which variable is a target of business data and a plurality of variables in the business data.
  • a reception unit that accepts designation as a variable, a statistical value calculation unit that calculates the amplitude of the time-series data related to the objective variable in a predetermined period, and within a predetermined period when the amplitude is larger than a predetermined amplitude threshold
  • a regression formula creation unit that creates a regression formula based on explanatory variables that are variables in business data correlated with the objective variable, and a display unit that displays the regression formula on a display device System.
  • an evaluation method for evaluating the effect of introducing a data analysis system including a reception step for accepting designation of which variable is a target variable among business data and a plurality of variables in the business data, and a time series regarding the target variable Statistical value calculation step for obtaining the amplitude of the data in a predetermined period, and if the amplitude is larger than the predetermined amplitude threshold, the objective variable in the predetermined period and the variable in the business data correlated with the objective variable
  • a regression formula creating step for creating a regression formula based on a certain explanatory variable and a display step for displaying the regression formula on a display device.
  • the present invention it becomes easier for the customer to confirm in advance the possibility that the objective variable will rise and the possibility that the measure can be implemented. This makes it easier to allocate computer resources based on information such as which objective variable the data analysis system analyzes and for which period the data analysis system analyzes. it can.
  • the data analysis system in this embodiment refers to a system that analyzes customer data, supports customer decision making, and directly controls plants and equipment.
  • An example is a system that uses machine learning and statistical analysis as the main data analysis methods.
  • the evaluation system in the present embodiment creates a regression equation within a period in which the fluctuation of the objective variable is large and extracts effective explanatory variables in order to determine in advance the effect of introducing the data analysis system. It is a system to do.
  • FIG. 1 is a diagram showing a specific example of the overall outline of the present embodiment. The configuration and data processing of the customer data center 10 and the external data analysis service center 11 will be described.
  • the customer data center 10 designates which of the plurality of variables of the business data 101 and business data collected from the workplaces such as stores, branches, and sites as the target variable, and receives from the data analysis service center
  • the system terminal 102 or the like that displays the evaluation result. Specification of the objective variable by the business data 101 and the system terminal 102 is transmitted to the data analysis service center 11 via the network 103.
  • the data analysis service center 11 analyzes the data collected from the customer data center 10 via the input / output unit 104 and the data analysis system 105 that actually analyzes the data according to instructions from the customer data center.
  • the system is configured by an evaluation system 106 that evaluates the effect of introducing the data analysis system 105 and creates a regression equation.
  • the business data 101 and the objective variable designation transmitted from the customer data center 10 are respectively input to the reception unit 107 in the evaluation system 106 via the input / output unit 104.
  • 1 includes a reception unit 107, a data analysis unit 108, and a display unit 109.
  • the receiving unit 107 receives and aggregates the business data 101 and the designation of the objective variable made by the system terminal 102.
  • the objective variable is a variable designated from a plurality of variables in the business data 101, and is usually a business index that defines a desirable state such as a customer wanting to increase or decrease. Say. Examples of the objective variable include a management index such as sales, work productivity, and product performance variation.
  • the number of objective variables to be specified is not limited to one, and there may be a plurality of target variables. When a plurality of objective variables are specified, the data analysis unit 108 analyzes each objective variable.
  • Business data refers to any data related to the work at the customer's workplace. For example, data on employee work, data on communication between employees, employee ID / position / gender data, sales data, data on work efficiency such as the number of work, and the like.
  • FIG. 14 is a diagram showing an example of the business data 101.
  • the POS table 1401 is a table showing a part of the business data 101 related to the merchandise purchase of the customer of the retail store
  • the work table 1402 is a table showing a part of the business data 101 related to the work of the warehouse.
  • the POS table 1401 stores purchase price, purchase points, and customer unit price data for all products in the retail store and for each product displayed in the area A. For example, a retail store manager who is a customer for the data analysis service center designates the total value 1403 of the total purchase amount as an objective variable.
  • the work table 1402 stores a work ID in the warehouse, a worker ID, the number of work that is the number of work per minute, the start time of the work, and the product ID of the work target.
  • a warehouse manager who is a customer for the data service center designates the number of operations 1404 as an objective variable. Examples of the work here include packing work and picking work in a warehouse.
  • the data analysis unit 108 in FIG. 1 receives the business data 101 and the objective variable designation from the reception unit 107, and determines the introduction effect of the data analysis system 105 based on the fluctuation width of the time-series data regarding the objective variable. Then, based on the determination result and the determination result, the regression equation created based on the objective variable and the explanatory variable highly correlated with the objective variable is transmitted to the display unit 109.
  • the display unit 109 displays the introduction effect determination result and regression equation received from the data analysis unit 108 on the display device of the system terminal 102 via the input / output unit 104 and the network 103.
  • the data analysis service center 11 is configured outside the customer data center 10, but a system having the function of the data analysis service center 11 may be configured in the customer data center 10, as shown in FIG. It is not limited to the configuration.
  • a system having the function of the data analysis service center 11 may be configured in the customer data center 10, as shown in FIG. It is not limited to the configuration.
  • evaluation system 106 is described as being configured in the data analysis service center 11, it may be configured in an external service center.
  • the evaluation system of the present invention can be applied to customers of all types of business that collect data on business.
  • customer industries for example, logistics, financial, manufacturing, retail, medical, infrastructure, and the like are assumed.
  • FIG. 2 is a diagram illustrating an example of a hardware configuration that implements the evaluation system 106 according to the present embodiment.
  • the hardware configuration in the evaluation system 106 is realized using a computer system (computer), and includes at least one set of CPU 201, ROM 202, RAM 203, keyboard 204, display device 205, HDD 206, printer 207, mouse 208, and data bus 209. Composed.
  • ROM 202 stores an OS (operating system) of the evaluation system 106 and the like.
  • the RAM 203 stores software relating to the evaluation system 106 and a database (not shown) that stores threshold values such as a shake threshold value described later.
  • a keyboard 204 operates the CPU 201.
  • the HDD 206 stores input data and analysis data.
  • the display device 205 shows input data, analysis data, a process of data analysis, and the like.
  • a mouse 208 operates the CPU 201.
  • the data bus 209 communicates each data.
  • the CPU 201 executes software related to data analysis stored in the RAM 203, thereby realizing each function shown in FIG.
  • the hardware configuration in the data analysis system 105 is the same as that shown in FIG. 2, the details are omitted, but in the data analysis system 105, by executing software related to data analysis stored in the RAM 203, Each function in the data analysis system 105 to be described later can be realized.
  • FIG. 3 is an example of a configuration diagram of the evaluation system 106.
  • the data analysis unit 108 of the evaluation system 106 includes a statistical value calculation unit 301 and a regression equation creation unit 302.
  • the evaluation system 106 receives from the input / output unit 104 designation of customer business data 101 and objective variables from the system terminal 102 by the customer. Further, the statistical value calculation unit 301 calculates the amplitude of the data related to the objective variable, and the regression equation creation unit 302 determines the effect of introducing the data analysis system based on the amplitude calculated by the statistical value calculation unit 301. Based on the determination result, a regression equation is created using an objective variable and explanatory variables highly correlated with the objective variable. In addition, as will be described later with reference to FIG. 15, the regression equation creation unit 302 can also create a regression equation using a composite explanatory variable generated from a plurality of explanatory variables.
  • the accepting unit 107 receives the designation of the business data 101 and the objective variable as described above, and then transmits the designation of the business data 101 and the objective variable to the statistical value calculation unit 301.
  • the statistical value calculation unit 301 receives the customer business data 101 and the designation of the objective variable from the reception unit 107.
  • business data originally acquired by the customer may be received, or business data may be acquired by installing various sensors at the customer's workplace. By having various sensors installed, it is possible to evaluate the effect of introduction using business data that is more realistic, and the accuracy of the evaluation is improved.
  • the designation of the objective variable is illustrated as being performed from the customer data center 10, an arbitrary objective variable may be designated in the data analysis service center 11, and the statistical value calculation unit 301 may receive it.
  • Statistic value calculation section 301 extracts time series data related to the specified objective variable based on the received business data and the specification of the objective variable, and calculates the fluctuation width of the time series data related to the objective variable in a predetermined period.
  • Fig. 4 is a graph showing time-series data regarding objective variables in a graph. The calculation process of the statistical value calculation unit 301 will be described with reference to FIG.
  • the statistical value calculation unit 301 extracts data related to the objective variable designated by the system terminal 102 from a plurality of variables included in the business data.
  • the objective variable a and the objective variable b are extracted.
  • the statistical value calculation unit 301 creates time series data 401 for the objective variable a and time series data 402 for the objective variable b.
  • the data is illustrated as continuous time-series data, but discontinuous data may be used.
  • the continuous time series data is obtained by an existing curve generation technique such as spline curve generation.
  • the amplitude of the extracted time series data 401 and 402 during a predetermined period t1 to t2 is calculated.
  • those input from the outside of the statistical value calculation unit 301 may be used, or those determined arbitrarily by the statistical value calculation unit 301 may be used.
  • the entire period of data related to the objective variable acquired in the past may be t1 to t2, or if the business environment changes for some reason, the period from the change timing to the present is set as t1 to t2. Also good.
  • the total period of data related to the objective variable acquired in the past is t1 to t2, the introduction effect can be evaluated based on more data, and the accuracy of the evaluation result is improved.
  • the process may be performed in different periods.
  • the time-series data amplitude relating to the objective variable is a numerical value representing how much the objective variable has changed in the past, and represents the amplitude of the waveform of the time-series data.
  • the difference 403 between the average value a1 of the data and the maximum value of the data 403, the difference 404 between the average value of the data a1 and the minimum value of the data 404, the difference 406 between the maximum value of the data and the minimum value of the data, or the like can be considered.
  • the difference with the average value is taken in the above, the median can be considered similarly.
  • Time series data with a large fluctuation width means that the fluctuation has been large in the past. That is, an objective variable with such time series data represents that the numerical value changes on a daily or temporal basis, and it is highly likely that the objective variable has the potential to increase or decrease the numerical value. It can be said that it is easy to improve by introducing a data analysis system.
  • Patent Document 3 the idea is opposite to the technique of determining that the data of each set is worthy of analysis as the sum of the values of the sum of squared deviations of the data of each set is smaller. I want to be.
  • the statistical value calculation unit 301 transmits the calculated fluctuation width to the regression equation creation unit 302.
  • the target for calculating the amplitude is the time series data of the objective variable itself.
  • this fluctuation range a target of processing, it is possible to evaluate by just taking the maximum value and the minimum value of the operation evaluation index calculated based on the probability density of the result of operation status determination as in Patent Document 2. It is possible to evaluate the possibility of change of the objective variable itself that could not be made.
  • the regression equation creation unit 302 determines the effect of introducing the data analysis system 105 based on the fluctuation width of the data related to the objective variable received from the statistical value calculation unit 301, and creates a regression equation based on the determination result.
  • the time series data 401 relating to the objective variable “a” represents data having a large amplitude
  • the time series data 402 relating to the objective variable “b” represents data having a small amplitude
  • the regression equation creation unit 302 holds a shake width threshold value in advance. This fluctuation threshold is a threshold for determining the effect of introducing the data analysis system 105.
  • the regression formula creation unit 302 compares the swing width between t1 and t2 calculated by the swing width calculation unit 301 with the swing threshold, and if the swing width between t1 and t2 is larger than the swing threshold, data analysis is performed. It is determined that the system 105 has “large introduction effect”, and when the amplitude between t1 and t2 is less than the amplitude threshold, it is determined that “introduction effect is small”.
  • the swing threshold when the objective variable is sales 10% of the average value of daily sales is considered. For example, when the average daily sales of a certain store is 100 (10,000 yen), the swing width threshold may be set to 10 (10,000 yen). Although the threshold is determined based on the daily sales, the same can be considered for one week or one month.
  • the fluctuation threshold when the objective variable is productivity a value of 5% of the value indicating productivity can be considered.
  • the swing width threshold is set to 8 (pieces). The same applies to not only the number but also the probability such as the yield ratio in the manufacture of semiconductor elements.
  • the shake threshold may be input from outside the evaluation system 106.
  • the swing threshold may be set together with the system terminal 102 of the customer data center 10 specifying the objective variable.
  • the accepting unit 107 accepts the shake width threshold. In this way, it is possible to determine the introduction effect based on the increase in the objective variable requested by the customer.
  • a regression equation is created using an objective variable between t1 and t2 and an explanatory variable highly correlated with the objective variable.
  • the regression equation is created from the explanatory variable generated based on the business data 101 received from the customer data center 10 and the objective variable designated by the customer.
  • a method for generating the explanatory variable from the business data 101 a conventional technique may be used.
  • the explanatory variable may be generated by a generation logic set in advance as described in Patent Document 1. In these processes, since explanatory variables are generated only for the period determined to have a large effect of introducing the data analysis system, it is possible to reduce the load on the system due to a reduction in the processing amount.
  • the regression equation creation unit 302 examines the relationship between the specified objective variable and the generated explanatory variable that causes the specified target variable, and executes statistical calculation to clarify each relationship. Examples of statistical calculation here include regression analysis. By the statistical calculation here, it is possible to extract explanatory variables that are statistically highly correlated with respect to the designated target variable in a period in which the fluctuation range is large. Here, a regression equation is created based on the extracted explanatory variable and the specified objective variable. Note that the regression equation creation unit 302 can also create a regression equation using a composite explanatory variable generated from a plurality of explanatory variables, as will be described later with reference to FIG. As the regression equation, the following equation (1) can be considered.
  • the regression equation creation unit 302 transmits the determined introduction effect and the created regression equation information to the display unit 109.
  • the display unit 109 receives the introduction effect determination result and the created regression equation from the regression equation creation unit 302, converts the received determination result and regression equation into a format for display on the display device of the system terminal 102, and the input / output unit The data is transmitted to the system terminal 102 via 104. Moreover, you may display on the display apparatus 205 which the evaluation system 106 has.
  • FIG. 5 is an example of a diagram illustrating a processing flow of the evaluation system 106.
  • step 501 the reception unit 107 receives designation of the business data 101 and the objective variable and transmits it to the statistical value calculation unit 301.
  • step 502 the statistical value calculation unit 301 creates time series data of the objective variable, calculates the fluctuation width of the time series data in a predetermined period, and transmits it to the regression equation creation unit 302.
  • Steps 503 to 508 are processed by the regression equation creation unit 302. Through these processes, the introduction effect of the data analysis system 105 is determined for the specified objective variable, and a regression equation is created.
  • step 503 the fluctuation width of the time series data received by the statistical value calculation unit 301 is compared with a fluctuation width threshold held in advance.
  • step 503 If it is determined in step 503 that the data amplitude is greater than the amplitude threshold, it is determined that the data analysis system 105 has a “large introduction effect” (step 504), and the data amplitude is less than the amplitude threshold. Is determined to be “small introduction effect” (step 505).
  • step 506 an explanatory variable having a high correlation with the objective variable received by the regression equation creation unit 302 is extracted, and a regression equation is created from the objective variable and the extracted explanatory variable.
  • step 507 the regression effect creation unit 302 transmits the introduction effect determined as “small introduction effect” to the display unit 109.
  • step 508 the regression equation creation unit 302 transmits the introduction effect determined to be “large introduction effect” and the created regression equation to the display unit 109.
  • step 509 the display unit 109 converts the introduction effect and the regression equation into a format to be displayed on the display device of the system terminal 102, and transmits the result to the input / output unit 104.
  • FIG. 6 is a diagram showing an example of a screen displayed on the display device of the customer data center 10 through the display unit 109. This screen is generated by the display unit 109.
  • This display screen includes an objective variable designation area 601 and an introduction effect display area 602.
  • the objective variable designation area 601 is an area generated by the display unit 109 from before the processing by the evaluation system 106.
  • the introduction effect display area 602 is an area generated by the display unit 109 in response to the analysis result of the data analysis unit 108.
  • the objective variable designating area 601 is an area for designating an objective variable when performing the above-described introduction effect calculation process. This corresponds to the interface of the system terminal 102 in FIG.
  • a target variable 61 that the customer wants to increase and a swing threshold 62 for evaluating the effect of introducing the data analysis system are set.
  • “monthly sales of the entire store” is set as the objective variable
  • “300,000” is set as the threshold.
  • the data analysis unit 108 determines the introduction effect and creates a regression equation. If the determination result and the regression equation are created, the regression is performed.
  • the formula is displayed in the introduction effect display area 602.
  • the introduction effect display area 602 is an area for displaying the introduction effect and the regression equation, and relates to the determination result of the introduction effect determined by the regression equation creation unit 302, the regression equation, and the objective variable created by the statistical value calculation unit 301. Displays time-series data.
  • the determination result of the introduction effect by the determination in step 503 in FIG. 5 is displayed.
  • time series data display area 65 time series data generated by the statistical value calculation unit 301 in FIG. 3 is displayed.
  • the display is focused on the difference between the average value of the data and the maximum value of the data, but a different portion may be displayed as long as the fluctuation width of the time-series data can be understood.
  • Switching of the display of different portions is performed by a display graph switching button 66.
  • you may display together the information etc. which specify the objective variable, such as the name of objective variable, index ID, or the arrow etc. which represented the numerical value of the fluctuation threshold value by length as needed.
  • the regression formula created by the regression formula creation unit 302 is displayed in the regression formula display area 67.
  • the regression equation may be displayed as shown in the figure if it is determined as “large introduction effect” in the determination in step 503 in FIG. 5, but if it is determined as “low introduction effect”, nothing is displayed. It may not be displayed, or information such as “do not create regression equation due to small introduction effect” may be displayed.
  • the customer can confirm what explanatory variable has a high correlation with the objective variable in a period in which the objective variable may increase. Therefore, it is necessary to examine in advance the necessary amount of computer resources to determine the effect of introducing a data analysis system, which period of analysis should be performed, and which objective variable should be analyzed. Is possible.
  • the evaluation system includes the receiving unit 107 that receives designation of which variable is the objective variable among the business data and a plurality of variables in the business data, and predetermined time-series data regarding the objective variable.
  • the evaluation method relates to an acceptance step 501 for accepting designation of which variable is the objective variable among the business data and a plurality of variables in the business data, and the objective variable.
  • Statistical value calculation step 502 for obtaining the amplitude of the time series data in a predetermined period, and the objective variable in the predetermined period and the business data correlated with the objective variable when the amplitude is larger than the predetermined amplitude threshold
  • a regression equation creating step 506 for creating a regression equation based on an explanatory variable
  • a display step 509 for displaying the regression equation on a display device.
  • the statistical value calculation unit 301 receives the customer business data 101 from the reception unit 107 and the designation of the objective variable from the system terminal 102, calculates the variance value of the time series data regarding the objective variable, and transmits it to the regression equation creation unit 302. .
  • Fig. 7 is a drawing showing time-series data regarding objective variables in a graph. Processing of the statistical value calculation unit 301 and the regression equation creation unit 302 in this modification will be described with reference to FIG.
  • the statistical value calculation unit 301 extracts data related to the objective variable from a plurality of variables included in the business data 101.
  • the objective variable c, the objective variable d, and the objective variable e are extracted.
  • the time series data 701 of the objective variable c, the time series data 702 of the objective variable d, and the time series data 703 of the objective variable e are created.
  • the variance values of the extracted time series data 701, 702 and 703 for a predetermined period t3 to t4 are calculated.
  • the predetermined periods t3 to t4 are determined similarly to t1 to t2.
  • the calculation method of the variance value is performed using a conventional method.
  • the variance value is calculated using the following equation (2).
  • the time series data 701 and 702 in FIG. 7 show an example where the variance value is small, and the time series data 703 shows an example where the variance value is large.
  • the fluctuation width of the time series data of the objective variable c and the objective variable d for example, the difference 704 between the average value c1 and the maximum value of the time series data 701 and the difference 705 between the average value d1 and the minimum value 705 of the time series data 702 are both. It can be seen that the values are larger than the difference 706 between the average value e1 and the maximum value of the time series data 703 and the difference 707 between the average value e1 and the minimum value.
  • an objective variable having a plurality of peaks such as objective variable e, has one peak such as a change in objective variable c or objective variable d, and the other part is an objective variable that is constantly changing.
  • the inventor has conceived that the effect of introducing the data analysis system 105 is greater. The reason is as follows.
  • the objective variable having such time series data is difficult to improve depending on the measures presented by the data analysis system.
  • the objective variable such as the objective variable c or the objective variable d has a small change during the period excluding the non-stationary event, so that the range in which the value increases or decreases is small, and the potential for rising or falling due to the measure can be small. High nature.
  • an objective variable having a plurality of peaks such as objective variable e
  • an objective variable having a plurality of peaks can present a measure for raising the value of a portion having a small value, and the measure presented by the data analysis system Is likely to improve the objective variable.
  • the effect of introducing the data analysis system is not necessarily high.
  • time-series data that has multiple peaks in other periods, or time-series data that has a peak like the objective variable g and that fluctuates regularly it is between t8 and t9. Since time series data including a plurality of peaks has a large variance value, the evaluation of such data does not become small.
  • data indicating the change in the objective variable f for example, the sales amount for each day of the retail store is considered, and it is considered that an event has occurred in the vicinity of the store at the timing of rapid change.
  • data indicating the change in the objective variable g for example, the number of visitors to the toy store per month can be considered, and a plurality of rapidly rising months may be December. .
  • the objective variable indicating a change such as 801 includes a period having a plurality of peaks between t5 and t7. Therefore, the objective variable is improved by analyzing the data of this period and introducing measures. Is possible.
  • the objective variable indicating a change such as 802 includes a plurality of peaks between t8 and t9, it is possible to present an effective measure for the corresponding period by analyzing the data of these periods. is there. Therefore, it is preferable to use a variance value within a predetermined period that can reduce the influence of some sudden changes for evaluation.
  • the statistical value calculation unit 301 transmits the calculated variance value to the regression equation creation unit 302.
  • the regression formula creation unit 302 holds a variance threshold value in advance, compares the variance value between t3 and t4 calculated by the statistical value calculation unit 301 with the variance threshold value, and the variance value between t3 and t4 is greater than the variance threshold value. If it is larger, it is determined that the data analysis system 105 is “large introduction effect”, and if the variance between t3 and t4 is less than the dispersion threshold, it is determined that “introduction effect is small”.
  • the variance threshold may be input from the outside of the evaluation system 106 in the same manner as the deflection threshold.
  • the regression equation creation unit 302 uses the objective variable between t3 and t4 and the explanatory variable highly correlated with the objective variable when it is determined that “the introduction effect is large” by comparison with the dispersion threshold. Create The regression equation is created by the method described above.
  • the regression equation creation unit 302 transmits the determined introduction effect and the created regression equation information to the display unit 109.
  • processing flow of the evaluation system 106 according to the present modification is obtained by changing the swing range in FIG. 5 to the variance value, and does not have a different flow other than that, and thus the illustration is omitted.
  • the evaluation system includes a statistical value calculation unit 301 that obtains a variance value for a predetermined period of time series data, and a regression that creates a regression equation when the variance value is greater than a predetermined variance threshold. It has a formula creation unit 302.
  • This configuration makes it possible to more accurately evaluate the effect of introducing time-series data having one peak in a steady change as a whole.
  • FIG. 18 is an example of a configuration diagram of the evaluation system 106 including the regression equation creation unit 302 having the improvement value calculation unit 1801.
  • the data analysis unit 108 calculates a predicted improvement value of the objective variable by introducing the data analysis system 105 and transmits it to the display unit 109.
  • Statistic value calculation unit 301 extracts time series data related to the specified objective variable, and calculates a deviation value in all data in a predetermined period of the time series data related to the objective variable. Thereafter, all the calculated deviation values are transmitted to the improved value calculation unit 1801 in the regression equation creation unit 302.
  • the deviation value is calculated here as an example, the median value may be calculated, the difference value calculated from the median value, and the subsequent processing may be performed.
  • the improved value calculation unit 1801 accepts all deviation values, and calculates a predicted improved value indicating an expected improvement of the objective variable by introducing the data analysis system 105 based on the deviation values.
  • the predicted improvement value is a value indicating how much the value of the objective variable specified by the customer may be improved by the introduction of the data analysis system. For example, if the objective variable is a variable related to the amount of money such as sales amount and customer unit price, the predicted improvement value is calculated as an amount of “1,000” yen, etc., and the objective variable relates to productivity such as work efficiency and defect occurrence rate. In the case of variables, the prediction improvement value is calculated by a ratio or number such as “13” percent, “10”, and the like.
  • the improved value calculation unit 1801 calculates the maximum deviation value 403, the minimum deviation value 404, the sum 406 of the absolute value of the minimum deviation value and the maximum deviation value 406, or the standard deviation value between t1 and t2. Predictive improvement value. Alternatively, a deviation exceeding the fluctuation width threshold value may be extracted, and an average of the extracted deviations may be obtained as a prediction improvement value.
  • the maximum deviation value as the predicted improvement value, you can evaluate how much the objective variable specified by the customer may increase, and by setting the minimum deviation value as the predicted improvement value, It is possible to evaluate how much the target variable can be reduced. Further, by using the sum of the absolute value of the minimum deviation and the maximum value of the deviation or the value of the standard deviation as the predicted improvement value, the possibility of raising the average of the objective variable can be evaluated.
  • the standard deviation value is calculated using a conventional method.
  • the standard deviation value is calculated using the following equation (3).
  • the improvement value calculation unit 1801 transmits the calculated predicted improvement value to the display unit 109.
  • the display unit 109 performs conversion for displaying the received predicted improvement value on the display device of the system terminal 102 of the customer data center 10, and transmits it to the system terminal 102 via the input / output unit 104.
  • FIG. 9 is a diagram showing an example of a screen displaying the predicted improvement value.
  • the same parts as those in FIG. 6 are denoted by the same reference numerals, and the content to be displayed and the operation of each component for display are the same.
  • the data analysis unit 108 calculates the predicted improvement value and displays the result in the numerical value display area 91.
  • the numerical value display area 91 is an area for displaying the predicted improvement value, and displays the predicted improvement value calculated by the improvement value calculation unit 1801.
  • the calculated prediction improvement value may be displayed as it is, or may be displayed after rounding off.
  • the objective variable is productivity or the like, for example, “15% decrease / month” is displayed.
  • Information specifying the objective variable such as the objective variable name and index ID may be displayed together as necessary.
  • This numerical display area 91 makes it easy to confirm the effect of introducing the data analysis system 105.
  • the evaluation system 106 includes the statistical value calculation unit 301 for obtaining a difference from the deviation value or the median value in a predetermined period for the time series data, and the difference from the deviation value or the median value.
  • a prediction improvement value calculation unit 1801 that obtains a prediction improvement value that is a value indicating improvement of an objective variable based on the introduction of the data analysis system 105, and a display unit 109 that displays the prediction improvement value on a display device. .
  • This configuration makes it easy to calculate the expected increase in actual objective variables. This makes it possible for the customer side to make a comparison with the introduction cost, making it easier to consider the introduction of the data analysis system.
  • the predicted improvement value is calculated at a rate such as “13” percent increase. Providing customers in advance how much the cost of the workplace will be reduced by this increase makes it easier to compare with the cost of introducing a data analysis system on the customer side.
  • a modified example of calculating a reduction cost which is a cost to be reduced when the calculated prediction improvement value is increased or decreased by the prediction improvement value when the objective variable attribute is productivity, will be described.
  • the objective variable attribute is a monetary amount
  • the predicted improvement value obtained by the above calculation is displayed as it is in the numerical value display area 91.
  • the reduction cost is reduced. It is obtained and displayed in the numerical value display area 91.
  • the basic system configuration is the same as in FIG. 18, except for the following points.
  • the accepting unit 107 accepts the attribute of the designated objective variable.
  • the objective variable attribute is designated and transmitted to the accepting unit 107 via the input / output unit 104.
  • the attribute of the objective variable may be specified in the data analysis service center 11 and transmitted to the reception unit 107. This makes it possible to reduce the burden on the customer side.
  • the objective variable attribute indicates whether or not the objective variable is related to a monetary amount, and is broadly divided into “amount” and “other”.
  • the reception unit 107 transmits the received attribute of the objective variable to the improvement value calculation unit 1801.
  • the improvement value calculation unit 1801 receives the attribute of the objective variable from the reception unit 107 and calculates the reduction cost.
  • the improvement value calculation unit 1801 holds a cost calculation index in advance.
  • the cost calculation index is an index that represents how much the cost is reduced by the increase or decrease in the case where the attribute of the objective variable is other than the monetary value. For example, if the number of work per unit time per worker is the objective variable, the work time of the worker that can be reduced by increasing the number of work is calculated. The cost to be reduced is calculated by multiplying the operation time. Thus, the amount of labor cost reduction for the number of work per unit time is an example of the cost calculation index.
  • Other examples of the cost calculation index include an average purchase amount per store visitor, a return work cost per machine stop probability of 1%, and the like.
  • the receiving unit 107 receives the cost calculation index via the input / output unit 104 and transmits it to the improved value calculation unit 1801.
  • the improvement value calculation unit 1801 calculates a reduction cost based on the calculated predicted improvement value and the cost calculation index. For example, by multiplying the predicted improvement value by the cost calculation index, the predicted improvement value when the attribute of the objective variable is not related to the amount can be expressed by the amount. Thereafter, the calculated reduction cost is transmitted to the display unit 109.
  • FIG. 10 is an example of a diagram illustrating a flow of a reduction cost calculation process of the improvement value calculation unit 1801.
  • step 1001 the improvement value calculation unit 1801 receives a deviation value from the statistical value calculation unit 301 and calculates a predicted improvement value. Since specific processing has already been described, it will be omitted.
  • step 1002 the improvement value calculation unit 1801 receives the attribute of the objective variable from the reception unit 107.
  • step 1003 the improvement value calculation unit 1801 determines whether or not the attribute of the objective variable received in step 1002 relates to the amount. If it is determined that the attribute of the objective variable is related to the amount, the process proceeds to step 1004. If it is determined that the attribute of the objective variable is related to other than the amount, the process proceeds to step 1005.
  • step 1004 the improvement value calculation unit 1801 transmits the predicted improvement value calculated in step 1001 to the display unit 109.
  • step 1005 the improvement value calculation unit 1801 calculates a reduction cost based on the predicted improvement value and the cost calculation index calculated in step 1001.
  • step 1006 the improvement value calculation unit 1801 transmits the reduction cost calculated in step 1005 to the display unit 109.
  • the evaluation system 106 includes the receiving unit 107 that receives the attribute of the objective variable, and when the attribute of the objective variable is not related to the amount of money, the time series data increases or decreases by the predicted improvement value.
  • an improvement value calculation unit 1801 for obtaining a reduction cost, which is a cost to be reduced, and a display unit 109 for displaying the reduction cost are provided.
  • the data analysis system evaluates the introduction effect of the data analysis system and creates the regression equation.
  • the data analysis system converts the business data to the business data.
  • the contents of the evaluation system for determining the analysis period width to be analyzed will be described.
  • the basic system configuration is the same as that shown in FIG. 3 except for the following points.
  • FIG. 11 is an example of a configuration diagram of the evaluation system 106 according to the present embodiment.
  • the evaluation system 106 determines the analysis period width of the business data, and transmits the determined analysis period width to the data analysis system 105 in the data analysis service center 11.
  • the evaluation system 106 further includes a period candidate creation unit 1101 and a period determination unit 1102 in addition to the configuration of FIG.
  • the period candidate creation unit 1101 receives the introduction effect determination result and the time-series data regarding the objective variable from the data analysis unit 108. If the data analysis unit 108 determines that the introduction effect determination result is large, the data analysis system 105 creates a period candidate that is a period candidate for analysis, and transmits the period candidate to the period determination unit 1102.
  • the period determination unit 1102 receives time-series data and period candidates related to the objective variable from the period candidate creation unit 1101. Further, based on the shape of the time series data regarding the objective variable in the period candidate, it is determined whether or not the accepted period candidate is suitable for analysis. Further, when it is determined that the analysis is suitable, the analysis period width of the corresponding period candidate is transmitted to the data analysis system 105.
  • the analysis period width is the time length of the business data to be used when the data analysis system 105 sequentially analyzes the business data 101.
  • the period candidate is temporarily created when the analysis period width is determined. It is a candidate for the period.
  • the data analysis system 105 particularly analyzes the business data 101 within the analysis period width in order to extract an explanatory variable having a large influence on the objective variable or present a measure for increasing the objective variable. .
  • FIG. 12 shows the analysis period when the evaluation system 106 determines that the amplitude of the time-series data related to the objective variable for a predetermined period is larger than the amplitude threshold, that is, when the introduction effect of the data analysis system is high. It is the figure which showed the example of the flow of the process which determines a width
  • the processing in step 1201 is performed by the period candidate creation unit 1101, and the processing from steps 1202 to 1204 is performed by the period determination unit 1102.
  • the period candidate creation unit 1101 receives time series data related to the objective variable from the data analysis unit 108, and creates a certain period randomly selected from the time series data as a period candidate. Thereafter, time-series data regarding the period candidate and the objective variable is transmitted to the period determining unit 1102.
  • the period determining unit 1102 receives time series data related to the period candidate and the objective variable from the period candidate creating unit 1101, and whether the maximum and minimum points are included in the period candidate for the received time series data related to the objective variable. judge.
  • the existing method is used to determine whether the local maximum point and the local minimum point are included.
  • FIG. 13 is a diagram showing an example of time-series data of objective variables determined to have a large introduction effect by the regression equation creation unit 302.
  • the time series data 1301 is created by the statistical value calculation unit 301.
  • a period candidate including at least one maximum point and one minimum point is suitable for analysis by the data analysis system 105.
  • the value of the objective variable is large even if data analysis is performed using business data for only that period. It is difficult to distinguish between a state and a small state.
  • the period in which the maximum point 1302 and the minimum point 1303 are included, such as between t10 and t11, is easy to distinguish between a large state and a small state of the objective variable. The inventor came up with that.
  • the data analysis system can easily present a measure for increasing the value of the small portion of the objective variable near the minimum point between t10 and t11 to the value near the maximum point. Therefore, it was decided that the period including the maximum point 1302 and the minimum point 1303, such as between t10 and t11, was suitable for the analysis period width. By analyzing the business data during such a period, it becomes easier to extract explanatory variables that explain the increase and decrease of the objective variable, and the quality of the data analysis is improved.
  • step 1202 If it is determined in step 1202 that at least one local maximum point and local minimum point are not included in the period candidates, such as t11 to t12 and t12 to t13, the process returns to step 1201 to generate the period candidates. Repeat the process. When it is determined that at least one local maximum point and local minimum point are included in the period candidates as in t10 to t11, the process proceeds to step 1203.
  • the period determination unit 1102 determines the period candidate determined in step 1202 as including at least one local maximum point and local minimum point as the analysis period width.
  • step 1204 the analysis period width determined by the period determination unit 1102 is transmitted to the data analysis system 105.
  • the period candidate that is a candidate for the period in which the data analysis system 105 performs the analysis. If the period candidate creation unit 1101 that creates the period candidate and the time series data in the period candidate period include at least one local maximum point and local minimum point, the data analysis system 105 analyzes the period candidate. A period determining unit 1102 that determines the analysis period width is provided.
  • Example 1 in order to determine the effect of introducing the data analysis system, an evaluation system that determines the effect of introduction and creates a regression equation has been described. Next, an example of a data analysis system that actually performs analysis using customer business data is shown.
  • FIG. 15 is a diagram showing an outline of the present embodiment including the detailed configuration of the data analysis system 105. The configuration and data processing of the customer data center 10 and the external data analysis system 105 will be described.
  • the business data 101 is stored in the customer data center 10.
  • the business data 101 is stored separately as past data 101A and current day data 101B. If they are not stored separately, the acquisition unit 120A, which will be described later, distinguishes the past data 101A and the current day data 101B after receiving the business data 101.
  • the customer business system 110 includes a business instruction unit 110A that gives business instructions to employees on the customer side, and a performance management unit 110B that confirms the business results for the business instructions. These contents can be confirmed from the system terminal 102.
  • FIG. 15 shows an example of the configuration of each of the customer data center 10 and the data analysis system 105 of this embodiment, and the data analysis system 105 may be incorporated in the business system 110 in the customer data center 10. . With this configuration, it is possible to complete the exchange of customer business data within the customer data center.
  • the data analysis system 105 includes a case management unit 120, a model generation unit 130, and a variable optimization unit 140.
  • the project management unit 120 performs optimization processing of objective variables such as productivity or sales for each business using necessary data.
  • the case management unit 120 acquires the specification of the past data 101A and which variable is the target variable from the customer data center 10 by the acquisition unit 120A, and transmits it to the model generation unit 130.
  • the model generation unit 130 includes a past data acquisition unit 131, a composite explanatory variable generation unit 132, a variable generation logic database 133, a composite explanatory variable database 134, a statistical modeling unit 135, and a model database 136.
  • the past data acquisition unit 131 receives the specification of the past data 101A and the objective variable, and combines the past data 101A received by the composite explanatory variable generation unit 132 in a composite manner and converts it into a composite explanatory variable. After that, it is stored in the composite explanatory variable database 134.
  • a stored variable generation logic 133 is used to generate the composite explanatory variable.
  • the variable generation logic 133 can use a preset logic, but can also use a logic that is dynamically changed by interlocking with a result of achievement managed in the achievement management unit 110B.
  • a process of generating a new composite explanatory variable that can affect the target variable by combining the data in a composite manner is performed. For example, a large number of complex explanatory variables exceeding 1000 to 1 million are automatically generated.
  • an explanatory variable generation process by a set of three operators, a conditional operator, a target operator, and a calculation operator, as described in Patent Document 1 may be performed, or other existing methods may be used.
  • a composite explanatory variable may be generated.
  • complex explanatory variables include conditional variables for certain explanatory variables, such as whether the room temperature is 15 degrees or less, or when the room temperature is 15 degrees or less and the number of operations is greater than 100 per minute. And a variable indicating “room temperature ⁇ number of work” when the objective variable tends to be large.
  • the statistical modeling unit 135 statistically analyzes a variable affecting the variable (objective variable y) related to sales, productivity, etc. designated by the customer from the composite explanatory variables stored in the composite explanatory variable database 134. These are selected and their relationship is created as a model, and the created model is stored in the model database 136.
  • the model here may be a model that can represent the relationship between the objective variable and the composite explanatory variable, and examples thereof include a regression equation.
  • the data analysis system 105 may be configured to create a model by receiving data and setting information from the business system 110 via the web API. By adopting such a configuration, it becomes unnecessary to create the data analysis system 105 having complicated individual programs for each business, so that it is possible to perform data analysis even without specialized knowledge of optimization problems. Software development effort can be reduced.
  • the matter management unit 120 acquires the model generated by the model generation unit 130 and transmits the model to the variable optimization unit 140. Next, processing in the variable optimization unit 140 will be described.
  • the variable optimization unit 140 includes a day data acquisition unit 141, a model acquisition unit 142, a combination change / composite explanatory variable generation unit 143, and an objective variable evaluation unit 144.
  • the day data acquisition unit 141 receives the day data 101B from the customer data center 10. Further, the model acquisition unit 142 receives the model 136 from the case management unit 120. When the business data 101 of the customer data center 10 is not stored separately for the past data 101A and the current day data 101B, the current day data acquisition unit 141 receives the current day data distinguished from the case management unit 120.
  • the model 136 and the day data 101B acquired in this way are transmitted to the combination change / composite explanatory variable generation unit 143.
  • the combination change / composite explanatory variable generation unit 143 changes a combination pattern of business tasks, inputs actual day data, and calculates a composite explanatory variable corresponding to the change of the combination pattern.
  • the compound explanatory variables are calculated by inputting the data for the current day against the model created using the past data, it is possible to create a model that takes into account the influence on the objective variable for the day, and at the same time Since the complex explanatory variable corresponding to the change of task combination pattern is calculated, the combination pattern that the objective variable is improved by repeating the evaluation of the objective variable and the change of the business task combination pattern described later Can be extracted.
  • the objective variable evaluation unit 144 evaluates the objective variable corresponding to the combination of business tasks by using the composite explanatory variable calculated by the combination change / composite explanatory variable generation unit 143.
  • the objective variable evaluation unit 144 determines whether or not the objective variable corresponding to the combination of business tasks has been improved. If the objective variable evaluation unit 144 determines that the value of the objective variable has been improved, 110. If it is determined that the value of the objective variable has not improved, the business task combination is not changed. In this example, the business task is changed only once, but it is also possible to perform this process repeatedly. Examples of processing of the objective variable evaluation unit 144 that performs repetitive processing are shown in 145 to 147.
  • the combination pattern of business tasks is sequentially changed, and exploratory processing is repeatedly performed so that the objective variable is improved.
  • a combination of business tasks is changed, a composite explanatory variable is calculated, and whether the objective variable is improved is checked (146). If the variable is not improved, the previous combination is held without changing the combination (147).
  • FIG. 16 shows the improvement rate of the objective variable when the above search process is performed, and the vertical axis indicates how much the objective variable has improved since the start of the search process.
  • the objective variable in FIG. 16 is a variable with a smaller value such as work time, for example.
  • a threshold of 5% is set as the improvement threshold 1601 is shown. It is set as an end condition that the value is less than the improvement threshold 1601.
  • FIG. 17 shows a process for optimizing the picking process order (cart No.) in the distribution warehouse as an example of the business improved by the above search process.
  • the work time is optimized by replacing the picking cart No. and then evaluating the overall work time using the model 136 and repeating the process of holding the replacement or returning to the combination before the replacement.
  • Picking processing order can be obtained.
  • the total work time as the objective variable is expressed as a combination of the stay time (explanatory variable) on each shelf of the cart in the picking work area.
  • the combination of the objective variable and the explanatory variable here is an example, and may be a combination of the objective variable such as sales and the explanatory variable such as the staying time of the clerk in the staying area.
  • the data analysis system 105 includes the acquisition unit 120A that receives designation of which variable is the target variable among the business data and the plurality of variables in the business data, and the past of the business data.
  • Statistical modeling unit 135 that creates the relationship between the objective variable and multiple variables in the business data as a model, and the task pattern of the business based on the data of the day of the model and the business data
  • the objective variable evaluation unit 144 is provided that calculates the change of the objective variable when the variable is changed, and maintains the pattern change when the objective variable is improved.
  • the customer actually changes the task pattern of the business to determine the effect, and without automatically determining the task pattern of the best business on a trial and error basis, It is possible to calculate a task pattern.
  • the data analysis system 105 can handle only necessary variables determined by the evaluation system 106 from the past data 101A.
  • a defined analysis period width determined at 106 can be used. Therefore, necessary calculation resources can be estimated in advance, and computer resources and the like can be used effectively.

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Tourism & Hospitality (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

This evaluation system, which displays the effect of introducing a data analysis system, comprises: a reception unit which receives specification of operating data and specification of a variable that is among the plurality of variables included in the operating data, and that is to be used as a response variable; a statistical value calculation unit which calculates the magnitude of fluctuations in time-series data for the response variable as measured over a predetermined period of time; a regression expression creation unit which, if the calculated magnitude of fluctuations exceeds a predetermined fluctuation magnitude threshold value, creates a regression expression on the basis of values of the response variable and values of explanatory variables as measured during the predetermined period of time, said explanatory variables being correlated with the response variable and included in the operating data; and a display unit which displays the regression expression on a display device. This configuration facilitates determination of whether the value of a selected response variable is likely to increase, and whether a selected measure can be implemented, before a data analysis system is introduced.

Description

評価システム、評価方法およびデータ解析システムEvaluation system, evaluation method, and data analysis system
　本発明は、評価システム、評価方法およびデータ解析システムに関する。 The present invention relates to an evaluation system, an evaluation method, and a data analysis system.
　近年、ビッグデータと呼ばれる大量のデータを解析し、今まで人が勘と経験で行ってきた意思決定を支援するデータ解析システムが急速に発展してきており、そのような解析の結果を利用して顧客に対して施策を提示する技術も発展してきている。 In recent years, data analysis systems that analyze large amounts of data called big data and support decision making that humans have done with intuition and experience have developed rapidly, and the results of such analysis can be used. Technology to present measures to customers is also developing.
　このようなデータ解析システムの例として、特許文献１には、設定した目的変数に対して統計的に相関の高い説明変数を順に羅列する処理や、説明変数と目的変数の関係の回帰式を作成する処理を行うデータ解析システムが記載されている。また、特許文献１では、未知の施策導入のリスクを軽減するためのサービス効果計算処理を行っている。この処理により、施策導入によって顧客や従業員の行動が変わることが、利益等にどのように寄与したのかを、定量的に評価している。 As an example of such a data analysis system, Patent Document 1 creates a process for sequentially listing explanatory variables that are statistically highly correlated with a set objective variable, and a regression equation for the relationship between the explanatory variable and the objective variable. A data analysis system that performs processing is described. Moreover, in patent document 1, the service effect calculation process for reducing the risk of unknown measure introduction is performed. This process quantitatively evaluates how changes in the behavior of customers and employees due to the introduction of measures contributed to profits.
　特許文献２には、ある操業変数についての操業評価指標の最大値と最小値の差分を基に、その操業変数の値の変更により操業評価指標がどの程度変化するかを反映した操業影響度というものを求めている。その操業影響度が大きい操業変数を抽出することにより、操業状況の改善に有効な操業変数を選択している。 Patent Document 2 refers to the operational impact that reflects how much the operation evaluation index changes due to the change in the value of the operation variable based on the difference between the maximum value and the minimum value of the operation evaluation index for a certain operation variable. I'm looking for something. By extracting an operation variable having a large operation influence degree, an operation variable effective for improving the operation state is selected.
　さらに、特許文献３には、目的変数に関するデータを分割し、分割された２つの目的変数の集合のそれぞれについて偏差平方和から求めるまとまり度を計算し、このまとまり度の値が大きい場合に、それぞれの集合に属するデータ間の統計的有意差が大きいと判断している。 Further, Patent Document 3 divides the data relating to the objective variable, calculates the unity degree obtained from the sum of squares of deviation for each of the two divided sets of objective variables, and when the value of the unity degree is large, It is judged that there is a large statistically significant difference between data belonging to the set.
特開２０１４－８１７５０号公報JP 2014-81750 A 特開２０１３－１４０５４８号公報JP 2013-140548 A 特開２００８－１６００８号公報JP 2008-16008 A
　データ解析システム導入により予測される効果は、データ解析システム導入前には顧客にとって不明確な場合もあった。これは、目的とするアウトカム（目的変数）が上昇する可能性があるか否かを施策の実施前に判断することが困難であることに起因する。困難である理由について下記に示す。 The effects predicted by the introduction of the data analysis system were sometimes unclear for customers prior to the introduction of the data analysis system. This is because it is difficult to determine whether or not the target outcome (objective variable) may increase before the implementation of the measure. The reason why it is difficult will be described below.
　特許文献１においては、施策の導入の効果を評価するために、行動変化施策を実施し、一定期間後にサービス効果計算処理を行っている。しかし、この方法では、少なくとも一回はデータ解析システムが提示した施策を実施する必要があり、施策の実施の前にサービス効果を評価することは困難であった。 In Patent Document 1, in order to evaluate the effect of introduction of a measure, a behavior change measure is implemented, and a service effect calculation process is performed after a certain period of time. However, with this method, it is necessary to implement the measure presented by the data analysis system at least once, and it is difficult to evaluate the service effect before the implementation of the measure.
　特許文献２では、操業変数毎に、操業状況の良否判定結果の確率密度に基づいて算出した操業評価指標の最大値と最小値から操業影響度を求めて、操業状況の改善に有効な操業変数を求めている。しかしながら、この計算方法では、あくまで操業変数が操業評価指標に与える影響を求めることは出来るが、操業状況自体に改善する余地があるかどうかの評価をすることは難しかった。例えば、減風発生割合に影響を最も与える操業変数を抽出することは可能だが、データ解析システムの提示する施策の実行により、減風発生割合を低下させ、炉況の好調を維持出来るかどうかの直接的な判断は困難であった。そのため、特許文献２に記載の技術では目的変数（操業状況）自体が向上する可能性を有しているかの判断をすることが難しかった。 In Patent Document 2, for each operation variable, the operation variable is obtained from the maximum value and the minimum value of the operation evaluation index calculated based on the probability density of the determination result of the operation status, and the operation variable effective for improving the operation status is obtained. Seeking. However, with this calculation method, it is possible to determine the influence of the operation variable on the operation evaluation index, but it is difficult to evaluate whether there is room for improvement in the operation state itself. For example, it is possible to extract the operating variables that have the greatest effect on the rate of wind reduction, but whether the data analysis system can reduce the rate of wind reduction and maintain good furnace conditions by implementing the measures presented by the data analysis system. Direct judgment was difficult. For this reason, it has been difficult to determine whether the objective variable (operation status) itself has a possibility of improving with the technique described in Patent Document 2.
　さらに、特許文献３では、分割された目的変数の集合ごとに偏差平方和を求め、各集合の偏差平方和を足した値が小さい場合には、分割された２つの目的変数の集合間の統計的有意差が大きいと判断し、その分割により生じた２つの集合をもとにデータ解析を行っている。つまり、各集合内のデータの偏差平方和の値が小さくなればなるほど、そのような集合を生む分割点が解析に適していると判断している。しかしながら、この方法では、目的変数のデータを解析するための分割点を適切に判断することは出来るが、目的変数そのものに向上する余地があるかどうかの判断をすることは困難であった。 Further, in Patent Document 3, a sum of squared deviations is obtained for each set of divided objective variables, and if the sum of the squared deviations of each set is small, statistics between two divided sets of objective variables are obtained. The data analysis is performed based on the two sets generated by the division. That is, as the value of the sum of squared deviations of the data in each set becomes smaller, it is determined that the division points that generate such a set are more suitable for analysis. However, with this method, it is possible to appropriately determine the dividing point for analyzing the data of the objective variable, but it is difficult to determine whether there is room for improvement in the objective variable itself.
　このように、従来の技術では、目的変数を変動させる他の説明変数を効率的に見つけ出すことは検討されていたが、目的変数自体が上昇する可能性が高いかどうかの評価についての検討は行われていなかった。更に、目的変数の増減の評価が行われていなく、どの目的変数に着目して解析を行えば良いか判断することが困難であった。さらに、コンピュータ資源等のリソースをどのように準備するかの検討が容易ではなかった。 As described above, in the conventional technique, it has been studied to efficiently find other explanatory variables that change the objective variable. However, the evaluation of whether the objective variable itself is likely to rise is not conducted. It wasn't. Furthermore, since the increase / decrease of the objective variable has not been evaluated, it has been difficult to determine which objective variable should be analyzed. Further, it has not been easy to examine how to prepare resources such as computer resources.
　また、特許文献１などに記載のデータ解析システムは、多数の説明変数を生成してその中から目的変数と相関の高い説明変数を自動抽出するため、人手によるデータの解析では着目が難しい説明変数を抽出することが特徴の一つである。従って、別の課題として、仮に上昇の可能性のある目的変数に着目して解析または目的変数の上昇の可能性のある期間について解析を行っていたとしても、顧客にとって制御困難または不可能な説明変数が抽出される可能性もあった。 In addition, since the data analysis system described in Patent Document 1 generates a large number of explanatory variables and automatically extracts explanatory variables having a high correlation with the objective variable from them, explanatory variables that are difficult to focus on in manually analyzing data Is one of the features. Therefore, as another issue, even if the analysis is performed focusing on the objective variable that is likely to rise, or the period during which the objective variable is likely to rise is analyzed, it is difficult or impossible for the customer to explain. There was also the possibility of extracting variables.
　以上を踏まえ、本願は、データ解析システムが提案した施策によって、顧客が指定したパフォーマンスを示す指標である目的変数に改善効果をもたらすかどうかをデータ解析システムの導入前に評価し、その評価の高い目的変数と、その評価の高い期間において目的変数と相関の高い説明変数等の情報を表示することをより容易にする技術を提供することを目的とする。 Based on the above, this application evaluates whether the measures proposed by the data analysis system have an improvement effect on the objective variable that is an indicator of the performance specified by the customer before introducing the data analysis system. It is an object of the present invention to provide a technique that makes it easier to display information such as an objective variable and an explanatory variable having a high correlation with the objective variable during a period of high evaluation.
　前記課題を解決するための手段のうち代表的なものを例示すれば、データ解析システムの導入効果を表示する評価システムであって、業務データおよび業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける受付部と、目的変数に関する時系列データの所定の期間における振れ幅を求める統計値計算部と、振れ幅が所定の振れ幅閾値より大きい場合に、所定の期間内の目的変数、および、目的変数と相関のある業務データ内の変数である説明変数を基に回帰式を作成する回帰式作成部と、回帰式を表示装置に表示する表示部と、を有する評価システムが挙げられる。 An example of a representative means for solving the above problem is an evaluation system for displaying the effect of introducing a data analysis system, and which variable is a target of business data and a plurality of variables in the business data. A reception unit that accepts designation as a variable, a statistical value calculation unit that calculates the amplitude of the time-series data related to the objective variable in a predetermined period, and within a predetermined period when the amplitude is larger than a predetermined amplitude threshold And a regression formula creation unit that creates a regression formula based on explanatory variables that are variables in business data correlated with the objective variable, and a display unit that displays the regression formula on a display device System.
　また、データ解析システムの導入効果を評価する評価方法であって、業務データおよび業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける受付ステップと、目的変数に関する時系列データの所定の期間における振れ幅を求める統計値計算ステップと、振れ幅が所定の振れ幅閾値より大きい場合に、所定の期間内の目的変数、および目的変数と相関のある業務データ内の変数である説明変数を基に回帰式を作成する回帰式作成ステップと、回帰式を表示装置に表示する表示ステップとを有する評価方法が挙げられる。 Also, an evaluation method for evaluating the effect of introducing a data analysis system, including a reception step for accepting designation of which variable is a target variable among business data and a plurality of variables in the business data, and a time series regarding the target variable Statistical value calculation step for obtaining the amplitude of the data in a predetermined period, and if the amplitude is larger than the predetermined amplitude threshold, the objective variable in the predetermined period and the variable in the business data correlated with the objective variable There is an evaluation method having a regression formula creating step for creating a regression formula based on a certain explanatory variable and a display step for displaying the regression formula on a display device.
　本発明によれば、顧客が事前に目的変数が上昇する可能性及び施策の実施できる可能性を確認することがより容易になる。これにより、どの目的変数に対してデータ解析システムの解析を行うか、どの期間のデータに対してデータ解析システムの解析を行うか等の情報に基づいたコンピュータリソースの配分もより容易に行うことができる。 According to the present invention, it becomes easier for the customer to confirm in advance the possibility that the objective variable will rise and the possibility that the measure can be implemented. This makes it easier to allocate computer resources based on information such as which objective variable the data analysis system analyzes and for which period the data analysis system analyzes. it can.
本発明の全体概要の例を示した図である。It is the figure which showed the example of the whole outline | summary of this invention. 評価システムのハードウェア構成の例を示した図である。It is the figure which showed the example of the hardware constitutions of an evaluation system. 評価システムにおけるシステム構成の例を示した図である。It is the figure which showed the example of the system configuration | structure in an evaluation system. 目的変数の時系列データのグラフの例を示した図である。It is the figure which showed the example of the graph of the time series data of an objective variable. 評価システムの処理のフローの例を示した図である。It is the figure which showed the example of the flow of a process of an evaluation system. 表示装置に表示される画面の例を示した図である。It is the figure which showed the example of the screen displayed on a display apparatus. 目的変数の時系列データのグラフの例を示した図である。It is the figure which showed the example of the graph of the time series data of an objective variable. 目的変数の時系列データのグラフの例を示した図である。It is the figure which showed the example of the graph of the time series data of an objective variable. 表示装置に表示される画面の例を示した図である。It is the figure which showed the example of the screen displayed on a display apparatus. 回帰式作成部の処理のフローの例を示した図である。It is the figure which showed the example of the flow of a process of a regression type preparation part. 評価システムにおけるシステム構成の例を示した図である。It is the figure which showed the example of the system configuration | structure in an evaluation system. 解析期間幅の決定処理のフローの例を示した図である。It is the figure which showed the example of the flow of the determination process of an analysis period width. 目的変数の時系列データのグラフの例を示した図である。It is the figure which showed the example of the graph of the time series data of an objective variable. 業務データの例を示した図である。It is the figure which showed the example of business data. データ解析システムにおけるシステム構成の例を示した図である。It is the figure which showed the example of the system configuration | structure in a data analysis system. 目的変数の改善を表すグラフの例を示した図である。It is the figure which showed the example of the graph showing the improvement of an objective variable. 目的変数の改善を表す表の例を示した図である。It is the figure which showed the example of the table showing the improvement of an objective variable. 評価システムにおけるシステム構成の例を示した図である。It is the figure which showed the example of the system configuration | structure in an evaluation system.
　本実施例では、目的変数に関するデータの振れ幅を計算して、データの振れ幅が大きい期間内の回帰式を表示する評価システムの例を説明する。 In this embodiment, an example of an evaluation system that calculates the fluctuation width of the data related to the objective variable and displays a regression equation within a period when the fluctuation width of the data is large will be described.
　本明細書中では、評価システムとデータ解析システムを区別して記載している。本実施例におけるデータ解析システムとは、顧客のデータを解析し、顧客の意思決定を支援、プラントや機器を直接制御するシステムのことをいう。例として、機械学習、統計分析を主なデータ解析の手法としているシステムが挙げられる。データ解析システムでは、業務に関するデータから、例えば１万個を超える非常に多数の説明変数を生成してデータの解析を行う。それに対し、本実施例における評価システムとは、データ解析システムの導入の効果を事前に判定するために、目的変数の振れ幅が大きい期間内の回帰式を作成し、効果的な説明変数を抽出するためのシステムのことである。 In this specification, the evaluation system and the data analysis system are described separately. The data analysis system in this embodiment refers to a system that analyzes customer data, supports customer decision making, and directly controls plants and equipment. An example is a system that uses machine learning and statistical analysis as the main data analysis methods. In the data analysis system, a very large number of explanatory variables exceeding, for example, 10,000 are generated from business-related data to analyze the data. On the other hand, the evaluation system in the present embodiment creates a regression equation within a period in which the fluctuation of the objective variable is large and extracts effective explanatory variables in order to determine in advance the effect of introducing the data analysis system. It is a system to do.
　以下、図面を参照しながら説明する。 Hereinafter, description will be made with reference to the drawings.
　図１は、本実施例の全体概要の具体例を示した図である。顧客データセンタ１０および外部のデータ解析サービスセンタ１１のそれぞれの構成およびデータ処理について説明する。 FIG. 1 is a diagram showing a specific example of the overall outline of the present embodiment. The configuration and data processing of the customer data center 10 and the external data analysis service center 11 will be described.
　顧客データセンタ１０は、店舗、支店、現場等の職場から集められた業務データ１０１および業務データの内の複数の変数のうちどの変数を目的変数とするかを指定し、データ解析サービスセンタから受け付けた評価結果を表示するシステム端末１０２等から構成される。業務データ１０１およびシステム端末１０２による目的変数の指定は、ネットワーク１０３経由でデータ分析サービスセンタ１１に送信される。 The customer data center 10 designates which of the plurality of variables of the business data 101 and business data collected from the workplaces such as stores, branches, and sites as the target variable, and receives from the data analysis service center The system terminal 102 or the like that displays the evaluation result. Specification of the objective variable by the business data 101 and the system terminal 102 is transmitted to the data analysis service center 11 via the network 103.
　データ解析サービスセンタ１１は、顧客データセンタ側の指示により、実際にデータの解析を行うデータ解析システム１０５、および、顧客データセンタ１０から入出力部１０４を経由して収集されたデータを分析してデータ解析システム１０５の導入効果を評価し、回帰式を作成する評価システム１０６等により構成される。顧客データセンタ１０から送信された業務データ１０１および目的変数の指定はそれぞれ、入出力部１０４を経由して評価システム１０６内の受付部１０７に入力される。 The data analysis service center 11 analyzes the data collected from the customer data center 10 via the input / output unit 104 and the data analysis system 105 that actually analyzes the data according to instructions from the customer data center. The system is configured by an evaluation system 106 that evaluates the effect of introducing the data analysis system 105 and creates a regression equation. The business data 101 and the objective variable designation transmitted from the customer data center 10 are respectively input to the reception unit 107 in the evaluation system 106 via the input / output unit 104.
　図１の評価システム１０６は、受付部１０７、データ分析部１０８および表示部１０９を備えている。 1 includes a reception unit 107, a data analysis unit 108, and a display unit 109.
　受付部１０７は、業務データ１０１およびシステム端末１０２により行われた目的変数の指定を受け付けて集約する。業務データ１０１を受け付けた際に、評価システム１０６の処理に適切なデータ形式に変換してもよい。目的変数とは、業務データ１０１中の複数の変数の中から指定された変数のことであり、通常は、顧客が上昇させたい、減少させたい等、望ましい状態を規定する業務上の指標のことをいう。目的変数としては、例えば、売上等の経営指標や作業の生産性、製品の性能ばらつきなどが挙げられる。指定される目的変数は１つとは限らず複数個あっても構わない。複数の目的変数が指定された場合には、それぞれの目的変数についてデータ分析部１０８により分析を行う。 The receiving unit 107 receives and aggregates the business data 101 and the designation of the objective variable made by the system terminal 102. When the business data 101 is received, it may be converted into a data format suitable for the processing of the evaluation system 106. The objective variable is a variable designated from a plurality of variables in the business data 101, and is usually a business index that defines a desirable state such as a customer wanting to increase or decrease. Say. Examples of the objective variable include a management index such as sales, work productivity, and product performance variation. The number of objective variables to be specified is not limited to one, and there may be a plurality of target variables. When a plurality of objective variables are specified, the data analysis unit 108 analyzes each objective variable.
　業務データとは、顧客の職場での業務に関する任意のデータのことをいう。例えば、従業員の作業に関するデータや、従業員間でのコミュニケーションに関するデータ、従業員のＩＤ・職位・性別のデータ、売上データ、または、作業個数等の作業効率に関するデータなどが挙げられる。 Business data refers to any data related to the work at the customer's workplace. For example, data on employee work, data on communication between employees, employee ID / position / gender data, sales data, data on work efficiency such as the number of work, and the like.
　図１４は、業務データ１０１の例を示した図である。ＰＯＳテーブル１４０１は小売店の来客の商品購入に関する業務データ１０１の一部を示した表であり、作業テーブル１４０２は倉庫の作業に関する業務データ１０１の一部を示した表である。 FIG. 14 is a diagram showing an example of the business data 101. The POS table 1401 is a table showing a part of the business data 101 related to the merchandise purchase of the customer of the retail store, and the work table 1402 is a table showing a part of the business data 101 related to the work of the warehouse.
　ＰＯＳテーブル１４０１には、購入金額、購入点数、客単価のデータが小売店の全商品及びエリアＡに陳列されている商品ごとに格納されている。例えば、データ解析サービスセンタにとっての顧客である小売店管理者は、購入金額の全商品合計の値１４０３を目的変数として指定する。 The POS table 1401 stores purchase price, purchase points, and customer unit price data for all products in the retail store and for each product displayed in the area A. For example, a retail store manager who is a customer for the data analysis service center designates the total value 1403 of the total purchase amount as an objective variable.
　作業テーブル１４０２には、倉庫内の作業ＩＤ、作業者ＩＤ、毎分の作業個数である作業数、作業の開始時刻、作業対象の商品ＩＤが格納されている。例えば、データサービスセンタにとっての顧客である倉庫管理者は、作業数１４０４を目的変数として指定する。ここでの作業は、例えば倉庫での梱包作業、ピッキング作業等が挙げられる。 The work table 1402 stores a work ID in the warehouse, a worker ID, the number of work that is the number of work per minute, the start time of the work, and the product ID of the work target. For example, a warehouse manager who is a customer for the data service center designates the number of operations 1404 as an objective variable. Examples of the work here include packing work and picking work in a warehouse.
　次に、図１のデータ分析部１０８は、受付部１０７から業務データ１０１および目的変数の指定を受け付けて、目的変数に関する時系列のデータの振れ幅を基にデータ解析システム１０５の導入効果を判定し、判定結果および判定結果に基づいて、目的変数およびその目的変数と相関の高い説明変数を基に作成した回帰式を表示部１０９に送信する。 Next, the data analysis unit 108 in FIG. 1 receives the business data 101 and the objective variable designation from the reception unit 107, and determines the introduction effect of the data analysis system 105 based on the fluctuation width of the time-series data regarding the objective variable. Then, based on the determination result and the determination result, the regression equation created based on the objective variable and the explanatory variable highly correlated with the objective variable is transmitted to the display unit 109.
　表示部１０９では、データ分析部１０８から受け付けた導入効果の判定結果および回帰式を、入出力部１０４およびネットワーク１０３を介してシステム端末１０２の表示装置に表示する。 The display unit 109 displays the introduction effect determination result and regression equation received from the data analysis unit 108 on the display device of the system terminal 102 via the input / output unit 104 and the network 103.
　図１では、データ解析サービスセンタ１１は顧客データセンタ１０の外部に構成されているが、データ解析サービスセンタ１１の機能を有するシステムが顧客データセンタ１０内に構成されていてもよく、図１の構成に限らない。データ解析サービスセンタ１１の機能を有するシステムを顧客データセンタ１０内に構成することにより、データの処理をすべて顧客データセンタ内で行うことができるので、顧客情報の保護のためにデータ変換をする必要がなく、データの処理速度向上につながる。 In FIG. 1, the data analysis service center 11 is configured outside the customer data center 10, but a system having the function of the data analysis service center 11 may be configured in the customer data center 10, as shown in FIG. It is not limited to the configuration. By configuring the system having the function of the data analysis service center 11 in the customer data center 10, all data processing can be performed in the customer data center. Therefore, it is necessary to convert data for protecting customer information. This leads to improved data processing speed.
　更に、評価システム１０６がデータ解析サービスセンタ１１内に構成されているように記載されているが、外部のサービスセンタに構成されていても構わない。 Furthermore, although the evaluation system 106 is described as being configured in the data analysis service center 11, it may be configured in an external service center.
　また、顧客の業種に関しては、特に限定する必要はなく、業務に関してデータを集約しているあらゆる業種の顧客に対して本発明の評価システムを適用することができる。顧客の業種としては例えば、物流業、金融業、製造業、小売業、医療業およびインフラ業等が想定される。 Further, there is no particular limitation on the customer's business type, and the evaluation system of the present invention can be applied to customers of all types of business that collect data on business. As customer industries, for example, logistics, financial, manufacturing, retail, medical, infrastructure, and the like are assumed.
　図２は、本実施例における評価システム１０６を実現するハードウェア構成の一例を示す図である。 FIG. 2 is a diagram illustrating an example of a hardware configuration that implements the evaluation system 106 according to the present embodiment.
　評価システム１０６におけるハードウェア構成は、コンピュータシステム(計算機)を用いて実現され、少なくとも１組の、ＣＰＵ２０１、ＲＯＭ２０２、ＲＡＭ２０３、キーボード２０４、表示装置２０５、ＨＤＤ２０６、プリンタ２０７、マウス２０８およびデータバス２０９から構成される。 The hardware configuration in the evaluation system 106 is realized using a computer system (computer), and includes at least one set of CPU 201, ROM 202, RAM 203, keyboard 204, display device 205, HDD 206, printer 207, mouse 208, and data bus 209. Composed.
　ＲＯＭ２０２は、評価システム１０６のＯＳ(オペレーティングシステム)等を格納する。ＲＡＭ２０３は、評価システム１０６に関するソフトウェアおよび後述する振れ幅閾値等の各閾値を格納するデータベース（図示せず）を格納する。キーボード２０４は、ＣＰＵ２０１を操作する。ＨＤＤ２０６は、入力データや分析データを格納する。表示装置２０５は、入力データ、分析データまたはデータ分析の処理の過程等を示す。マウス２０８は、ＣＰＵ２０１を操作する。データバス２０９は、各々のデータを通信する。 ROM 202 stores an OS (operating system) of the evaluation system 106 and the like. The RAM 203 stores software relating to the evaluation system 106 and a database (not shown) that stores threshold values such as a shake threshold value described later. A keyboard 204 operates the CPU 201. The HDD 206 stores input data and analysis data. The display device 205 shows input data, analysis data, a process of data analysis, and the like. A mouse 208 operates the CPU 201. The data bus 209 communicates each data.
　評価システム１０６において、ＣＰＵ２０１で、ＲＡＭ２０３に格納されたデータ分析に関するソフトウェアを実行することで、図１に示した各機能を実現することが出来る。 In the evaluation system 106, the CPU 201 executes software related to data analysis stored in the RAM 203, thereby realizing each function shown in FIG.
　更に、データ解析システム１０５におけるハードウェア構成も図２に示したものと同一であるため、詳細は省略するが、データ解析システム１０５において、ＲＡＭ２０３に格納されたデータ解析に関するソフトウェアを実行することで、後述するデータ解析システム１０５における各機能を実現することが可能となる。 Further, since the hardware configuration in the data analysis system 105 is the same as that shown in FIG. 2, the details are omitted, but in the data analysis system 105, by executing software related to data analysis stored in the RAM 203, Each function in the data analysis system 105 to be described later can be realized.
　図３は、評価システム１０６の構成図の例である。評価システム１０６のデータ分析部１０８は、統計値計算部３０１および回帰式作成部３０２を備えている。評価システム１０６は、入出力部１０４から、顧客の業務データ１０１および顧客によるシステム端末１０２からの目的変数の指定を受け付ける。また、統計値計算部３０１により目的変数に関するデータの振れ幅を計算し、回帰式作成部３０２により、統計値計算部３０１により計算された振れ幅を基にデータ解析システム導入の効果を判定し、判定結果に基づいて目的変数および目的変数と相関の高い説明変数により回帰式を作成する。また、回帰式作成部３０２においては、図１５において後述するように、複数の説明変数から生成された複合説明変数を用いて、回帰式を作成することもできる。 FIG. 3 is an example of a configuration diagram of the evaluation system 106. The data analysis unit 108 of the evaluation system 106 includes a statistical value calculation unit 301 and a regression equation creation unit 302. The evaluation system 106 receives from the input / output unit 104 designation of customer business data 101 and objective variables from the system terminal 102 by the customer. Further, the statistical value calculation unit 301 calculates the amplitude of the data related to the objective variable, and the regression equation creation unit 302 determines the effect of introducing the data analysis system based on the amplitude calculated by the statistical value calculation unit 301. Based on the determination result, a regression equation is created using an objective variable and explanatory variables highly correlated with the objective variable. In addition, as will be described later with reference to FIG. 15, the regression equation creation unit 302 can also create a regression equation using a composite explanatory variable generated from a plurality of explanatory variables.
　受付部１０７は、前述のように業務データ１０１および目的変数の指定を受け付けた後、統計値計算部３０１に業務データ１０１および目的変数の指定を送信する。 The accepting unit 107 receives the designation of the business data 101 and the objective variable as described above, and then transmits the designation of the business data 101 and the objective variable to the statistical value calculation unit 301.
　統計値計算部３０１は、受付部１０７から顧客の業務データ１０１および目的変数の指定を受け付ける。 The statistical value calculation unit 301 receives the customer business data 101 and the designation of the objective variable from the reception unit 107.
　業務データの取得方法としては、もともと顧客側が取得していた業務データを受け取ってもよいし、顧客の職場に各種センサを設置してもらい業務データを取得してもよい。各種センサを設置してもらうことにより、より実態に即した業務データを用いて導入効果の評価をすることができ、評価の精度が向上する。 As a method of acquiring business data, business data originally acquired by the customer may be received, or business data may be acquired by installing various sensors at the customer's workplace. By having various sensors installed, it is possible to evaluate the effect of introduction using business data that is more realistic, and the accuracy of the evaluation is improved.
　目的変数の指定は顧客データセンタ１０から行われているように図示されているが、データ解析サービスセンタ１１内で任意の目的変数を指定し、それを統計値計算部３０１が受けとってもよい。 Although the designation of the objective variable is illustrated as being performed from the customer data center 10, an arbitrary objective variable may be designated in the data analysis service center 11, and the statistical value calculation unit 301 may receive it.
　統計値計算部３０１は、受け付けた業務データおよび目的変数の指定を基に、指定された目的変数に関する時系列データを抽出し、目的変数に関する時系列データの所定の期間における振れ幅を計算する。 Statistic value calculation section 301 extracts time series data related to the specified objective variable based on the received business data and the specification of the objective variable, and calculates the fluctuation width of the time series data related to the objective variable in a predetermined period.
　図４は、目的変数に関する時系列データをグラフにして表した図面である。統計値計算部３０１の計算処理について、図４を用いて説明する。 Fig. 4 is a graph showing time-series data regarding objective variables in a graph. The calculation process of the statistical value calculation unit 301 will be described with reference to FIG.
　まず、統計値計算部３０１は、業務データに含まれる複数の変数の中から、システム端末１０２により指定された目的変数に関するデータを抽出する。ここでは、目的変数ａおよび目的変数ｂが抽出されたとする。 First, the statistical value calculation unit 301 extracts data related to the objective variable designated by the system terminal 102 from a plurality of variables included in the business data. Here, it is assumed that the objective variable a and the objective variable b are extracted.
　続いて、統計値計算部３０１は、目的変数ａの時系列データ４０１および目的変数ｂの時系列データ４０２を作成する。図４では、連続的な時系列データとして図示しているが、非連続な点のデータでも構わない。連続的な時系列データは、例えば、スプライン曲線生成などの既存の曲線生成技術によって行う。 Subsequently, the statistical value calculation unit 301 creates time series data 401 for the objective variable a and time series data 402 for the objective variable b. In FIG. 4, the data is illustrated as continuous time-series data, but discontinuous data may be used. The continuous time series data is obtained by an existing curve generation technique such as spline curve generation.
　次に、抽出した時系列データ４０１および４０２の所定の期間ｔ１～ｔ２の間の振れ幅を計算する。ここで、所定の期間ｔ１～ｔ２は、統計値計算部３０１の外部から入力されたものを用いても良いし、統計値計算部３０１が任意に決定したものでもよい。例えば、過去に取得した目的変数に関するデータの全期間をｔ１～ｔ２としても良いし、何らかの事情で業務の環境が変化した場合には変化したタイミングから現在までの期間をｔ１～ｔ２として設定しても良い。過去に取得した目的変数に関するデータの全期間をｔ１～ｔ２とした場合には、より多くのデータに基づいた導入効果の評価をすることができ、評価結果の精度が向上する。また、業務の環境が変化したタイミングから現在までの期間をｔ１～ｔ２とした場合には、現在の業務環境に則した導入効果の評価をすることができる。なお、ここでは時系列データ４０１および４０２の所定の期間を同じ期間としているが、異なる期間で処理を行っても構わない。 Next, the amplitude of the extracted time series data 401 and 402 during a predetermined period t1 to t2 is calculated. Here, for the predetermined periods t1 to t2, those input from the outside of the statistical value calculation unit 301 may be used, or those determined arbitrarily by the statistical value calculation unit 301 may be used. For example, the entire period of data related to the objective variable acquired in the past may be t1 to t2, or if the business environment changes for some reason, the period from the change timing to the present is set as t1 to t2. Also good. When the total period of data related to the objective variable acquired in the past is t1 to t2, the introduction effect can be evaluated based on more data, and the accuracy of the evaluation result is improved. In addition, when the period from the timing when the business environment changes to the present time is t1 to t2, it is possible to evaluate the introduction effect according to the current business environment. Although the predetermined period of the time series data 401 and 402 is the same period here, the process may be performed in different periods.
　目的変数に関する時系列データの振れ幅とは、目的変数が過去どの程度変化しているかを表した数値であり、時系列データの波形の振れ幅を表している。例えば、データの平均値ａ１とデータの最大値との差４０３、データの平均値ａ１とデータの最小値の差４０４、または、データの最大値と最小値の差４０６等が振れ幅として考えられる。また、上記では平均値との差をとっているが、中央値でも同様に考えられる。 The time-series data amplitude relating to the objective variable is a numerical value representing how much the objective variable has changed in the past, and represents the amplitude of the waveform of the time-series data. For example, the difference 403 between the average value a1 of the data and the maximum value of the data 403, the difference 404 between the average value of the data a1 and the minimum value of the data 404, the difference 406 between the maximum value of the data and the minimum value of the data, or the like can be considered. . Moreover, although the difference with the average value is taken in the above, the median can be considered similarly.
　振れ幅が大きい時系列データとは、過去に変動が大きかったことを表している。すなわち、そのような時系列データを持つ目的変数は、日常的にまたは時間的に数値が変化していることを表しており、その数値の上昇または下降のポテンシャルを持っている可能性が高く、データ解析システムの導入により改善がしやすいものであるということがいえる。この点で、特許文献３のように、各集合のデータの偏差平方和の値の和が小さくなればなるほど各集合のデータは解析に値すると判断する技術とは逆の思想であることに留意されたい。 Time series data with a large fluctuation width means that the fluctuation has been large in the past. That is, an objective variable with such time series data represents that the numerical value changes on a daily or temporal basis, and it is highly likely that the objective variable has the potential to increase or decrease the numerical value. It can be said that it is easy to improve by introducing a data analysis system. In this regard, it is noted that, as in Patent Document 3, the idea is opposite to the technique of determining that the data of each set is worthy of analysis as the sum of the values of the sum of squared deviations of the data of each set is smaller. I want to be.
　最後に、統計値計算部３０１は、計算した振れ幅を回帰式作成部３０２に送信する。 Finally, the statistical value calculation unit 301 transmits the calculated fluctuation width to the regression equation creation unit 302.
　ここで、振れ幅を計算する対象が目的変数の時系列データそのものであることに留意されたい。この振れ幅を処理の対象とすることにより、特許文献２のように操業状況の良否判定結果の確率密度に基づいて算出した操業評価指標の最大値と最小値を取るだけでは、評価することが出来なかった目的変数自体の変化の可能性を評価することが可能となる。 Note that the target for calculating the amplitude is the time series data of the objective variable itself. By making this fluctuation range a target of processing, it is possible to evaluate by just taking the maximum value and the minimum value of the operation evaluation index calculated based on the probability density of the result of operation status determination as in Patent Document 2. It is possible to evaluate the possibility of change of the objective variable itself that could not be made.
　回帰式作成部３０２は、統計値計算部３０１から受け付けた目的変数に関するデータの振れ幅を基に、データ解析システム１０５導入の効果を判定し、判定結果に基づき回帰式を作成する。 The regression equation creation unit 302 determines the effect of introducing the data analysis system 105 based on the fluctuation width of the data related to the objective variable received from the statistical value calculation unit 301, and creates a regression equation based on the determination result.
　図４の目的変数ａに関しての時系列データ４０１は振れ幅が大きいデータ、目的変数ｂに関しての時系列データ４０２は振れ幅が小さいデータを表したものである。 4, the time series data 401 relating to the objective variable “a” represents data having a large amplitude, and the time series data 402 relating to the objective variable “b” represents data having a small amplitude.
　回帰式作成部３０２は、予め振れ幅閾値を保持している。この振れ幅閾値は、データ解析システム１０５導入の効果を判定する閾値である。回帰式作成部３０２は、振れ幅計算部３０１が計算したｔ１～ｔ２間の振れ幅と振れ幅閾値を比較し、ｔ１～ｔ２間の振れ幅が振れ幅閾値よりも大きい場合には、データ解析システム１０５の「導入効果大」と判定し、ｔ１～ｔ２間の振れ幅が振れ幅閾値未満の場合には、「導入効果小」と判定する。 The regression equation creation unit 302 holds a shake width threshold value in advance. This fluctuation threshold is a threshold for determining the effect of introducing the data analysis system 105. The regression formula creation unit 302 compares the swing width between t1 and t2 calculated by the swing width calculation unit 301 with the swing threshold, and if the swing width between t1 and t2 is larger than the swing threshold, data analysis is performed. It is determined that the system 105 has “large introduction effect”, and when the amplitude between t1 and t2 is less than the amplitude threshold, it is determined that “introduction effect is small”.
　目的変数が売上の場合の振れ幅閾値として、１日の売上の平均値の１０パーセントの数値が考えられる。例えば、ある店舗の１日の売上の平均が１００（万円）であった場合には、振れ幅閾値は１０（万円）と設定してもよい。１日の売上で閾値を決めたが、１週間でも１ヶ月でも同様に考えられる。 * As the swing threshold when the objective variable is sales, 10% of the average value of daily sales is considered. For example, when the average daily sales of a certain store is 100 (10,000 yen), the swing width threshold may be set to 10 (10,000 yen). Although the threshold is determined based on the daily sales, the same can be considered for one week or one month.
　また、目的変数が生産性の場合の振れ幅閾値として、生産性を示す数値の５パーセントの数値が考えられる。例えば、任意の従業員のある物品のピッキング作業の１時間あたりの個数が１５０（個）であった場合には、振れ幅閾値は８（個）と設定される。個数だけでなく、半導体素子の製造における歩留まり割合などの確率でも同様に考えられる。 Also, as the fluctuation threshold when the objective variable is productivity, a value of 5% of the value indicating productivity can be considered. For example, when the number of picking operations for an article with an arbitrary employee per hour is 150 (pieces), the swing width threshold is set to 8 (pieces). The same applies to not only the number but also the probability such as the yield ratio in the manufacture of semiconductor elements.
　振れ幅閾値は、評価システム１０６の外部から入力されてもよい。例えば、顧客データセンタ１０のシステム端末１０２が目的変数を指定するのと合わせて、振れ幅閾値を設定するようにしても構わない。この場合には、受付部１０７が振れ幅閾値を受け付ける。このようにすることで、顧客が要求する目的変数の上昇分を基に導入効果を判定することが可能となる。 The shake threshold may be input from outside the evaluation system 106. For example, the swing threshold may be set together with the system terminal 102 of the customer data center 10 specifying the objective variable. In this case, the accepting unit 107 accepts the shake width threshold. In this way, it is possible to determine the introduction effect based on the increase in the objective variable requested by the customer.
　次に、回帰式作成部３０２は、指定された目的変数について、ｔ１～ｔ２間の振れ幅と振れ幅閾値との比較から、データ解析システム１０５の「導入効果大」と判定された場合に、ｔ１～ｔ２間の目的変数およびその目的変数と相関の高い説明変数を用いて回帰式を作成する。回帰式は、顧客データセンタ１０から受け付けた業務データ１０１を基に生成した説明変数と顧客に指定された目的変数から作成される。業務データ１０１から説明変数を生成する方法は、従来の技術を用いればよく、例えば、特許文献１に記載されているような事前に設定される生成ロジックによって、生成してもよい。これらの処理においては、データ解析システムの導入効果大と判定された期間についてのみ説明変数を生成していることから、処理量の低減によるシステムへの負荷を減らすことが可能となる。 Next, when the regression equation creation unit 302 determines that the specified objective variable is “large introduction effect” of the data analysis system 105 based on the comparison between the swing width between t1 and t2 and the swing threshold, A regression equation is created using an objective variable between t1 and t2 and an explanatory variable highly correlated with the objective variable. The regression equation is created from the explanatory variable generated based on the business data 101 received from the customer data center 10 and the objective variable designated by the customer. As a method for generating the explanatory variable from the business data 101, a conventional technique may be used. For example, the explanatory variable may be generated by a generation logic set in advance as described in Patent Document 1. In these processes, since explanatory variables are generated only for the period determined to have a large effect of introducing the data analysis system, it is possible to reduce the load on the system due to a reduction in the processing amount.
　具体的には、回帰式作成部３０２は、指定された目的変数と、その要因となる生成された説明変数との関係を調べて、それぞれの関係を明らかにする統計計算を実行する。ここでの統計計算としては、例えば回帰分析等が挙げられる。ここでの統計計算により、指定された目的変数に対して、その振れ幅の大きい期間において統計的に相関の高い説明変数を抽出することが可能となる。ここで、抽出した説明変数と指定された目的変数を基に回帰式を作成する。なお、回帰式作成部３０２においては、図１５において後述するように、複数の説明変数から生成された複合説明変数を用いて、回帰式を作成することもできる。
回帰式としては、以下のような式（１）が考えられる。 Specifically, the regression equation creation unit 302 examines the relationship between the specified objective variable and the generated explanatory variable that causes the specified target variable, and executes statistical calculation to clarify each relationship. Examples of statistical calculation here include regression analysis. By the statistical calculation here, it is possible to extract explanatory variables that are statistically highly correlated with respect to the designated target variable in a period in which the fluctuation range is large. Here, a regression equation is created based on the extracted explanatory variable and the specified objective variable. Note that the regression equation creation unit 302 can also create a regression equation using a composite explanatory variable generated from a plurality of explanatory variables, as will be described later with reference to FIG.
As the regression equation, the following equation (1) can be considered.
Figure JPOXMLDOC01-appb-M000001
 　その後、回帰式作成部３０２は、表示部１０９に判定した導入効果および作成した回帰式の情報を送信する。
Figure JPOXMLDOC01-appb-M000001
 Thereafter, the regression equation creation unit 302 transmits the determined introduction effect and the created regression equation information to the display unit 109.
　表示部１０９は、回帰式作成部３０２から導入効果の判定結果および作成された回帰式を受け付け、受け付けた判定結果および回帰式をシステム端末１０２の表示装置に表示させる形式に変換し、入出力部１０４を介してシステム端末１０２に送信する。また、評価システム１０６が持つ表示装置２０５に表示させても構わない。 The display unit 109 receives the introduction effect determination result and the created regression equation from the regression equation creation unit 302, converts the received determination result and regression equation into a format for display on the display device of the system terminal 102, and the input / output unit The data is transmitted to the system terminal 102 via 104. Moreover, you may display on the display apparatus 205 which the evaluation system 106 has.
　図５は、評価システム１０６の処理のフローを示した図の例である。 FIG. 5 is an example of a diagram illustrating a processing flow of the evaluation system 106.
　ステップ５０１では、受付部１０７により業務データ１０１および目的変数の指定を受け付け、統計値計算部３０１に送信する。 In step 501, the reception unit 107 receives designation of the business data 101 and the objective variable and transmits it to the statistical value calculation unit 301.
　ステップ５０２では、統計値計算部３０１により目的変数の時系列データを作成し、所定の期間における時系列データの振れ幅を計算し、回帰式作成部３０２に送信する。 In step 502, the statistical value calculation unit 301 creates time series data of the objective variable, calculates the fluctuation width of the time series data in a predetermined period, and transmits it to the regression equation creation unit 302.
　ステップ５０３～ステップ５０８は、回帰式作成部３０２によって処理される。これらの処理により、指定された目的変数に対して、データ解析システム１０５の導入効果を判定し、回帰式を作成する。 Steps 503 to 508 are processed by the regression equation creation unit 302. Through these processes, the introduction effect of the data analysis system 105 is determined for the specified objective variable, and a regression equation is created.
　ステップ５０３では、統計値計算部３０１が受け付けた時系列データの振れ幅と予め保持している振れ幅閾値とを比較する。 In step 503, the fluctuation width of the time series data received by the statistical value calculation unit 301 is compared with a fluctuation width threshold held in advance.
　ステップ５０３にて、データの振れ幅が振れ幅閾値より大きいと判定された場合には、データ解析システム１０５の「導入効果大」と判定され（ステップ５０４）、データの振れ幅が振れ幅閾値未満と判定された場合には、「導入効果小」と判定される（ステップ５０５）。 If it is determined in step 503 that the data amplitude is greater than the amplitude threshold, it is determined that the data analysis system 105 has a “large introduction effect” (step 504), and the data amplitude is less than the amplitude threshold. Is determined to be “small introduction effect” (step 505).
　ステップ５０６では、回帰式作成部３０２により受け付けた目的変数と相関の高い説明変数を抽出し、目的変数および抽出された説明変数から回帰式を作成する。 In step 506, an explanatory variable having a high correlation with the objective variable received by the regression equation creation unit 302 is extracted, and a regression equation is created from the objective variable and the extracted explanatory variable.
　ステップ５０７では、回帰式作成部３０２にて、「導入効果小」と判定された導入効果を表示部１０９に送信する。 In step 507, the regression effect creation unit 302 transmits the introduction effect determined as “small introduction effect” to the display unit 109.
　ステップ５０８では、回帰式作成部３０２にて、「導入効果大」と判定された導入効果及び作成された回帰式を表示部１０９に送信する。 In step 508, the regression equation creation unit 302 transmits the introduction effect determined to be “large introduction effect” and the created regression equation to the display unit 109.
　ステップ５０９では、表示部１０９により、導入効果および回帰式をシステム端末１０２の表示装置に表示する形式に変換し、その結果を入出力部１０４に送信する。 In step 509, the display unit 109 converts the introduction effect and the regression equation into a format to be displayed on the display device of the system terminal 102, and transmits the result to the input / output unit 104.
　図６は、表示部１０９を通して顧客データセンタ１０の表示装置に表示される画面の例を示した図である。この画面は、表示部１０９によって生成される。 FIG. 6 is a diagram showing an example of a screen displayed on the display device of the customer data center 10 through the display unit 109. This screen is generated by the display unit 109.
　本表示画面は、目的変数指定エリア６０１、導入効果表示エリア６０２から構成される。ここで、目的変数指定エリア６０１は、評価システム１０６による処理の前から、表示部１０９によって生成されているエリアである。これに対し導入効果表示エリア６０２は、データ分析部１０８の分析結果を受けて、表示部１０９が生成するエリアである。 This display screen includes an objective variable designation area 601 and an introduction effect display area 602. Here, the objective variable designation area 601 is an area generated by the display unit 109 from before the processing by the evaluation system 106. On the other hand, the introduction effect display area 602 is an area generated by the display unit 109 in response to the analysis result of the data analysis unit 108.
　目的変数指定エリア６０１は、上記の導入効果の計算処理を行う際の目的変数を指定するエリアである。これは、図１のシステム端末１０２のインタフェースに相当する。ここで、顧客が上昇させたい目的変数６１とデータ解析システムの導入効果を評価するための振れ幅閾値６２を設定する。ここでは、目的変数として、「店舗全体の月単位の売上」が設定され、閾値として、「３００，０００」が設定されている。目的変数と振れ幅閾値が設定され、計算実行ボタン６３が押されると、データ分析部１０８は導入効果の判定および回帰式の作成を行い、その判定結果および回帰式を作成した場合にはその回帰式を導入効果表示エリア６０２に表示する。 The objective variable designating area 601 is an area for designating an objective variable when performing the above-described introduction effect calculation process. This corresponds to the interface of the system terminal 102 in FIG. Here, a target variable 61 that the customer wants to increase and a swing threshold 62 for evaluating the effect of introducing the data analysis system are set. Here, “monthly sales of the entire store” is set as the objective variable, and “300,000” is set as the threshold. When the objective variable and the amplitude threshold are set and the calculation execution button 63 is pressed, the data analysis unit 108 determines the introduction effect and creates a regression equation. If the determination result and the regression equation are created, the regression is performed. The formula is displayed in the introduction effect display area 602.
　導入効果表示エリア６０２は、導入効果および回帰式を表示するエリアであり、回帰式作成部３０２により判定された導入効果の判定結果や回帰式、統計値計算部３０１にて作成された目的変数に関する時系列データ等を表示する。 The introduction effect display area 602 is an area for displaying the introduction effect and the regression equation, and relates to the determination result of the introduction effect determined by the regression equation creation unit 302, the regression equation, and the objective variable created by the statistical value calculation unit 301. Displays time-series data.
　結果表示エリア６４には、図５のステップ５０３の判定による導入効果の大小の判定結果が表示されている。 In the result display area 64, the determination result of the introduction effect by the determination in step 503 in FIG. 5 is displayed.
　また、時系列データ表示エリア６５には、図３の統計値計算部３０１により生成された時系列データが表示されている。ここでは、データの平均値とデータの最大値の差の部分に着目して表示させているが、時系列データの振れ幅が分かる図であれば異なる部分を表示させても構わない。異なる部分の表示の切り替えは、表示グラフ切替ボタン６６により実施される。また、必要に応じて目的変数の名称、指標ＩＤ等の目的変数を特定する情報または振れ幅閾値の数値を長さで表した矢印等を合わせて表示してもよい。 In the time series data display area 65, time series data generated by the statistical value calculation unit 301 in FIG. 3 is displayed. Here, the display is focused on the difference between the average value of the data and the maximum value of the data, but a different portion may be displayed as long as the fluctuation width of the time-series data can be understood. Switching of the display of different portions is performed by a display graph switching button 66. Moreover, you may display together the information etc. which specify the objective variable, such as the name of objective variable, index ID, or the arrow etc. which represented the numerical value of the fluctuation threshold value by length as needed.
　回帰式表示エリア６７には、回帰式作成部３０２によって作成された回帰式が表示されている。回帰式は、図５のステップ５０３による判定で「導入効果大」と判定された場合には、図のように表示すれば良いが、「導入効果小」と判定された場合には、何も表示しなくてもよいし、「導入効果小のため、回帰式の作成をしない」等といった情報を表示してもよい。この回帰式を確認することで、顧客は、目的変数の上昇の可能性のある期間における、その目的変数と相関の高い説明変数が何かを確認することができる。そのため、データ解析システムの導入の効果判断や、どの期間の解析を行えばよいか、どの目的変数に対して解析を行えばよいか、がわかるのでのコンピュータリソースの必要量についても事前検討することが可能となる。 The regression formula created by the regression formula creation unit 302 is displayed in the regression formula display area 67. The regression equation may be displayed as shown in the figure if it is determined as “large introduction effect” in the determination in step 503 in FIG. 5, but if it is determined as “low introduction effect”, nothing is displayed. It may not be displayed, or information such as “do not create regression equation due to small introduction effect” may be displayed. By confirming this regression equation, the customer can confirm what explanatory variable has a high correlation with the objective variable in a period in which the objective variable may increase. Therefore, it is necessary to examine in advance the necessary amount of computer resources to determine the effect of introducing a data analysis system, which period of analysis should be performed, and which objective variable should be analyzed. Is possible.
　このように、本実施例に係る評価システムは、業務データおよび業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける受付部１０７と、目的変数に関する時系列データの所定の期間内における振れ幅を求める統計値計算部３０１と、振れ幅が所定の振れ幅閾値より大きい場合に、所定の期間内の目的変数、および、目的変数と相関のある業務データ内の変数である説明変数を基に回帰式を作成する回帰式作成部３０２と、回帰式を表示装置に表示する表示部１０９を有することを特徴としている。 As described above, the evaluation system according to the present embodiment includes the receiving unit 107 that receives designation of which variable is the objective variable among the business data and a plurality of variables in the business data, and predetermined time-series data regarding the objective variable. A statistical value calculation unit 301 for obtaining a fluctuation within a predetermined period, and a target variable within a predetermined period and a variable in business data correlated with the objective variable when the fluctuation is greater than a predetermined fluctuation threshold. It is characterized by having a regression equation creation unit 302 that creates a regression equation based on an explanatory variable, and a display unit 109 that displays the regression equation on a display device.
　また、別の表現をすれば、本実施例に係る評価方法は、業務データおよび業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける受付ステップ５０１と、目的変数に関する時系列データの所定の期間における振れ幅を求める統計値計算ステップ５０２と、振れ幅が所定の振れ幅閾値より大きい場合に、所定の期間内の目的変数、および、目的変数と相関のある業務データ内の変数である説明変数を基に回帰式を作成する回帰式作成ステップ５０６と、前記回帰式を表示装置に表示する表示ステップ５０９を有することを特徴としている。 In other words, the evaluation method according to the present embodiment relates to an acceptance step 501 for accepting designation of which variable is the objective variable among the business data and a plurality of variables in the business data, and the objective variable. Statistical value calculation step 502 for obtaining the amplitude of the time series data in a predetermined period, and the objective variable in the predetermined period and the business data correlated with the objective variable when the amplitude is larger than the predetermined amplitude threshold A regression equation creating step 506 for creating a regression equation based on an explanatory variable, and a display step 509 for displaying the regression equation on a display device.
　係る構成により、データ解析システムを導入する前に、顧客が向上させたい目的変数に対してのデータ解析システム導入の効果を推定することが出来る。これにより、データ解析システムの導入効果の事前判断を容易にすることや、コンピュータ資源等のリソース準備についての検討を容易にすることが可能となる。 With such a configuration, it is possible to estimate the effect of introducing the data analysis system for the objective variable that the customer wants to improve before introducing the data analysis system. As a result, it is possible to facilitate the advance determination of the effect of introducing the data analysis system, and it is possible to facilitate the examination of resource preparation such as computer resources.
　＜分散計算を導入効果に用いる変形例＞
次に、目的変数に関するデータの分散値を計算して、データ解析システムの導入効果の判定を行う評価システムの例を説明する。基本的なシステム構成は図３と同じであるが、以下の点が相違する。 <Modified example using distributed calculation for introduction effect>
Next, an example of an evaluation system that calculates the variance value of data related to the objective variable and determines the effect of introducing the data analysis system will be described. The basic system configuration is the same as that shown in FIG. 3 except for the following points.
　統計値計算部３０１は受付部１０７から顧客の業務データ１０１およびシステム端末１０２からの目的変数の指定を受け付けて、目的変数に関する時系列データの分散値を計算し、回帰式作成部３０２に送信する。 The statistical value calculation unit 301 receives the customer business data 101 from the reception unit 107 and the designation of the objective variable from the system terminal 102, calculates the variance value of the time series data regarding the objective variable, and transmits it to the regression equation creation unit 302. .
　図７は、目的変数に関する時系列データをグラフにして表した図面である。本変形例における統計値計算部３０１および回帰式作成部３０２の処理について、図７を用いて説明する。 Fig. 7 is a drawing showing time-series data regarding objective variables in a graph. Processing of the statistical value calculation unit 301 and the regression equation creation unit 302 in this modification will be described with reference to FIG.
　まず、統計値計算部３０１は、業務データ１０１に含まれる複数の変数の中から、目的変数に関するデータを抽出する。ここでは、目的変数ｃ、目的変数ｄおよび目的変数ｅが抽出されたとする。続いて、目的変数ｃの時系列データ７０１、目的変数ｄの時系列データ７０２および目的変数ｅの時系列データ７０３を作成する。 First, the statistical value calculation unit 301 extracts data related to the objective variable from a plurality of variables included in the business data 101. Here, it is assumed that the objective variable c, the objective variable d, and the objective variable e are extracted. Subsequently, the time series data 701 of the objective variable c, the time series data 702 of the objective variable d, and the time series data 703 of the objective variable e are created.
　次に、抽出した時系列データ７０１、７０２および７０３の所定の期間ｔ３～ｔ４の分散値を計算する。ここで、所定の期間ｔ３～ｔ４は、ｔ１～ｔ２と同様に決められるものとする。また、分散値の計算方法は、従来の方法を用いて行う。例えば、以下の式（２）を用いて分散値を計算する。 Next, the variance values of the extracted time series data 701, 702 and 703 for a predetermined period t3 to t4 are calculated. Here, it is assumed that the predetermined periods t3 to t4 are determined similarly to t1 to t2. In addition, the calculation method of the variance value is performed using a conventional method. For example, the variance value is calculated using the following equation (2).
Figure JPOXMLDOC01-appb-M000002
 　図７の時系列データ７０１および７０２は、分散値が小さい例を、時系列データ７０３は、分散値が大きい例を示している。
Figure JPOXMLDOC01-appb-M000002
 The time series data 701 and 702 in FIG. 7 show an example where the variance value is small, and the time series data 703 shows an example where the variance value is large.
　ここで、目的変数ｃおよび目的変数ｄの時系列データの振れ幅、例えば時系列データ７０１の平均値ｃ１と最大値の差７０４及び時系列データ７０２の平均値ｄ１と最小値の差７０５は共に、時系列データ７０３の平均値ｅ１と最大値の差７０６及び平均値ｅ１と最小値の差７０７よりも大きい値となっていることがわかる。しかしながら、目的変数ｃまたは目的変数ｄの変化のようなピークを１つ持ち、他の部分は定常的な変動をしている目的変数よりも、目的変数ｅのような複数のピークを持つ目的変数の方が、データ解析システム１０５の導入による効果が大きいことに発明者は想到した。その理由は以下の通りである。 Here, the fluctuation width of the time series data of the objective variable c and the objective variable d, for example, the difference 704 between the average value c1 and the maximum value of the time series data 701 and the difference 705 between the average value d1 and the minimum value 705 of the time series data 702 are both. It can be seen that the values are larger than the difference 706 between the average value e1 and the maximum value of the time series data 703 and the difference 707 between the average value e1 and the minimum value. However, an objective variable having a plurality of peaks, such as objective variable e, has one peak such as a change in objective variable c or objective variable d, and the other part is an objective variable that is constantly changing. The inventor has conceived that the effect of introducing the data analysis system 105 is greater. The reason is as follows.
　目的変数ｃまたは目的変数ｄのようにピークを１つ持ち、他の部分は定常的な変動をしている目的変数は、そのピークが現れたタイミングで何らかの非定常的なイベントの影響を大きく受けていると考えられる。このような時系列データを持つ目的変数は、データ解析システムの提示する施策によっては改善が難しい。すなわち、目的変数ｃまたは目的変数ｄのような目的変数は、非定常的なイベントを除いた期間は変化が小さいので、その値の増減する幅が小さく、施策による上昇または下降のポテンシャルが小さい可能性が高い。これに対し、目的変数ｅのように複数のピークを持つ目的変数は、その値が小さい部分に関して、その値を上げるような施策の提示をすることが可能であり、データ解析システムの提示する施策によって目的変数が改善する可能性が高い。このように、データの振れ幅が大きい時系列データでも、必ずしもデータ解析システムの導入効果が高いとは限らない。 An objective variable that has one peak, such as objective variable c or objective variable d, and other parts that steadily fluctuate, is greatly affected by some non-stationary event at the time the peak appears. It is thought that. The objective variable having such time series data is difficult to improve depending on the measures presented by the data analysis system. In other words, the objective variable such as the objective variable c or the objective variable d has a small change during the period excluding the non-stationary event, so that the range in which the value increases or decreases is small, and the potential for rising or falling due to the measure can be small. High nature. On the other hand, an objective variable having a plurality of peaks, such as objective variable e, can present a measure for raising the value of a portion having a small value, and the measure presented by the data analysis system Is likely to improve the objective variable. As described above, even with time-series data having a large data fluctuation width, the effect of introducing the data analysis system is not necessarily high.
　ここで、分散とは、式（２）で表されるため、ｔ３～ｔ４が充分に大きい場合は目的変数ｃ及びｄに関する時系列データの分散値は小さくなる。そのため、これらの目的変数が指定されている場合には、導入効果は小さいと判定されることとなる。 Here, since the variance is expressed by the equation (2), when t3 to t4 are sufficiently large, the variance of the time series data regarding the objective variables c and d becomes small. Therefore, when these objective variables are designated, it is determined that the introduction effect is small.
　なお、図８の目的変数ｆのように一部の期間（ｔ７～ｔ６）でピークを１つ持ち、ｔ７～ｔ６内のピーク以外の他の部分が定常的な変動をしていたとしても、他の期間に複数のピークを持つような時系列データや、目的変数ｇのようにピークを持ち、その前後が定常的な変動をしている時系列データであっても、ｔ８～ｔ９間に複数のピークが含まれる時系列データは、分散値も大きくなるため、これらのようなデータの評価が小さくなることはない。目的変数ｆの変化を示すデータとしては、例えば、小売店の日ごとの売上額が考えられて、急激な変化をしているタイミングで店舗の付近でイベントが発生したと考えられる。また、目的変数ｇの変化を示すデータとしては、例えば、玩具店への月ごとの入場者数が考えられて、複数の急激な上昇をしている月は１２月である場合等が考えられる。 Even if there is one peak in some period (t7 to t6) as in the objective variable f in FIG. 8 and other parts other than the peak in t7 to t6 are constantly changing, Even if time-series data that has multiple peaks in other periods, or time-series data that has a peak like the objective variable g and that fluctuates regularly, it is between t8 and t9. Since time series data including a plurality of peaks has a large variance value, the evaluation of such data does not become small. As data indicating the change in the objective variable f, for example, the sales amount for each day of the retail store is considered, and it is considered that an event has occurred in the vicinity of the store at the timing of rapid change. Further, as data indicating the change in the objective variable g, for example, the number of visitors to the toy store per month can be considered, and a plurality of rapidly rising months may be December. .
　このように、８０１のような変化を示す目的変数は、ｔ５～ｔ７間等に複数のピークを持つ期間を含むため、この期間のデータを解析して施策を導入することにより、目的変数を改善することが可能である。また、８０２のような変化を示す目的変数はｔ８～ｔ９間等に複数のピークを含むため、これらの期間のデータの解析を行うことによって該当期間に対する効果的な施策を提示することが可能である。そこで、一部の突発的な変化による影響を少なくすることが可能な所定の期間内における分散値を評価に用いると良い。 In this way, the objective variable indicating a change such as 801 includes a period having a plurality of peaks between t5 and t7. Therefore, the objective variable is improved by analyzing the data of this period and introducing measures. Is possible. In addition, since the objective variable indicating a change such as 802 includes a plurality of peaks between t8 and t9, it is possible to present an effective measure for the corresponding period by analyzing the data of these periods. is there. Therefore, it is preferable to use a variance value within a predetermined period that can reduce the influence of some sudden changes for evaluation.
　最後に、統計値計算部３０１は、計算した分散値を回帰式作成部３０２に送信する。 Finally, the statistical value calculation unit 301 transmits the calculated variance value to the regression equation creation unit 302.
　回帰式作成部３０２は、予め分散閾値を保持しており、統計値計算部３０１が計算したｔ３～ｔ４間の分散値と分散閾値を比較し、ｔ３～ｔ４間の分散値が分散閾値よりも大きい場合には、データ解析システム１０５の「導入効果大」と判定し、ｔ３～ｔ４間の分散値が分散閾値未満の場合には、「導入効果小」と判定する。分散閾値は、振れ幅閾値と同様、評価システム１０６の外部から入力されても良い。 The regression formula creation unit 302 holds a variance threshold value in advance, compares the variance value between t3 and t4 calculated by the statistical value calculation unit 301 with the variance threshold value, and the variance value between t3 and t4 is greater than the variance threshold value. If it is larger, it is determined that the data analysis system 105 is “large introduction effect”, and if the variance between t3 and t4 is less than the dispersion threshold, it is determined that “introduction effect is small”. The variance threshold may be input from the outside of the evaluation system 106 in the same manner as the deflection threshold.
　続いて、回帰式作成部３０２は、分散閾値との比較で「導入効果大」と判定された場合に、ｔ３～ｔ４間の目的変数およびその目的変数と相関の高い説明変数を用いて回帰式を作成する。回帰式の作成は、上述した方法により行う。 Subsequently, the regression equation creation unit 302 uses the objective variable between t3 and t4 and the explanatory variable highly correlated with the objective variable when it is determined that “the introduction effect is large” by comparison with the dispersion threshold. Create The regression equation is created by the method described above.
　その後、回帰式作成部３０２は、表示部１０９に判定した導入効果および作成した回帰式の情報を送信する。 Thereafter, the regression equation creation unit 302 transmits the determined introduction effect and the created regression equation information to the display unit 109.
　なお、本変形例に係る評価システム１０６の処理フローについては、図５の振れ幅を分散値に変更したものであり、それ以外に異なるフローを有さないため図示を省略する。 Note that the processing flow of the evaluation system 106 according to the present modification is obtained by changing the swing range in FIG. 5 to the variance value, and does not have a different flow other than that, and thus the illustration is omitted.
　このように、本変形例に係る評価システムは、時系列データについて所定の期間における分散値を求める統計値計算部３０１と、分散値が所定の分散閾値より大きい場合に、回帰式を作成する回帰式作成部３０２を有することを特徴としている。 As described above, the evaluation system according to the present modification includes a statistical value calculation unit 301 that obtains a variance value for a predetermined period of time series data, and a regression that creates a regression equation when the variance value is greater than a predetermined variance threshold. It has a formula creation unit 302.
　係る構成により、全体が定常的な変化の中、１つのピークを持つ時系列データに関しての導入効果をより正確に評価できるようになる。 This configuration makes it possible to more accurately evaluate the effect of introducing time-series data having one peak in a steady change as a whole.
　＜導入効果を計算する変形例＞
導入効果の大小だけでなく、定量的な導入効果を事前に顧客に提示することは、定量的な導入効果を確認し、データ解析システム導入の費用と比較することが可能となるため、データ解析システムのスムースな導入につながる。 <Variation to calculate introduction effect>
Presenting quantitative introduction effects to customers in advance as well as the magnitude of the introduction effects makes it possible to confirm the quantitative introduction effects and compare it with the cost of introducing the data analysis system. This leads to a smooth introduction of the system.
　ここでは、生産性向上、費用削減、売上額上昇の値など、解析システムが提示する施策の実行による定量的な導入効果を事前に顧客に提示する内容に関して説明する。基本的なシステム構成は図３と同じであるが、以下の点で相違する。 Here, we will explain the contents to be presented to customers in advance with quantitative introduction effects due to the implementation of the measures presented by the analysis system, such as productivity improvement, cost reduction, and sales increase values. The basic system configuration is the same as that shown in FIG. 3 except for the following points.
　図１８は、改善値計算部１８０１を有する回帰式作成部３０２を含んだ評価システム１０６の構成図の例である。 FIG. 18 is an example of a configuration diagram of the evaluation system 106 including the regression equation creation unit 302 having the improvement value calculation unit 1801.
　データ分析部１０８は、データ解析システム１０５の導入による目的変数の予測改善値を計算して表示部１０９に送信する。 The data analysis unit 108 calculates a predicted improvement value of the objective variable by introducing the data analysis system 105 and transmits it to the display unit 109.
　統計値計算部３０１は、指定された目的変数に関する時系列データを抽出し、目的変数に関する時系列データの所定の期間の全てのデータにおける偏差の値を計算する。その後、計算したすべての偏差の値を回帰式作成部３０２内の改善値計算部１８０１に送信する。ここでは、例として偏差の値を計算しているが、中央値を計算して、中央値から差の値を計算し、その後の処理を行っても構わない。 Statistic value calculation unit 301 extracts time series data related to the specified objective variable, and calculates a deviation value in all data in a predetermined period of the time series data related to the objective variable. Thereafter, all the calculated deviation values are transmitted to the improved value calculation unit 1801 in the regression equation creation unit 302. Although the deviation value is calculated here as an example, the median value may be calculated, the difference value calculated from the median value, and the subsequent processing may be performed.
　改善値計算部１８０１は、すべての偏差の値を受け付けて、それを基にデータ解析システム１０５の導入による目的変数の改善の予想を示す予測改善値について計算する。 The improved value calculation unit 1801 accepts all deviation values, and calculates a predicted improved value indicating an expected improvement of the objective variable by introducing the data analysis system 105 based on the deviation values.
　予測改善値とは、データ解析システムの導入により顧客が指定した目的変数の値がどの程度改善する可能性があるかを示した値である。例えば、目的変数が売上額、客単価等の金額に関する変数の場合は、予測改善値は「１，０００」円等の金額として算出され、目的変数が作業効率、不具合発生割合等の生産性に関する変数の場合は、予測改善値は「１３」パーセント、「１０」個等の割合や個数で算出される。 The predicted improvement value is a value indicating how much the value of the objective variable specified by the customer may be improved by the introduction of the data analysis system. For example, if the objective variable is a variable related to the amount of money such as sales amount and customer unit price, the predicted improvement value is calculated as an amount of “1,000” yen, etc., and the objective variable relates to productivity such as work efficiency and defect occurrence rate. In the case of variables, the prediction improvement value is calculated by a ratio or number such as “13” percent, “10”, and the like.
　改善値計算部１８０１の計算方法について、図４を用いて説明する。改善値計算部１８０１は、偏差の最大値４０３、偏差の最小値４０４、偏差の最小値の絶対値と偏差の最大値の和４０６、または、ｔ１～ｔ２間の標準偏差の値を計算し、予測改善値とする。または、振れ幅閾値を超える偏差を抽出し、抽出された偏差の平均を求めて予測改善値としても構わない。 The calculation method of the improvement value calculation unit 1801 will be described with reference to FIG. The improved value calculation unit 1801 calculates the maximum deviation value 403, the minimum deviation value 404, the sum 406 of the absolute value of the minimum deviation value and the maximum deviation value 406, or the standard deviation value between t1 and t2. Predictive improvement value. Alternatively, a deviation exceeding the fluctuation width threshold value may be extracted, and an average of the extracted deviations may be obtained as a prediction improvement value.
　偏差の最大値を予測改善値とすることで、顧客が指定した目的変数についてどの程度上昇する可能性があるかの評価をすることができ、偏差の最小値を予測改善値とすることで、目的変数についてどの程度減少を抑えられる可能性があるかの評価をすることができる。また、偏差の最小値の絶対値と偏差の最大値の和や標準偏差の値を予測改善値とすることで、目的変数の平均を底上げできる可能性を評価することができる。 By using the maximum deviation value as the predicted improvement value, you can evaluate how much the objective variable specified by the customer may increase, and by setting the minimum deviation value as the predicted improvement value, It is possible to evaluate how much the target variable can be reduced. Further, by using the sum of the absolute value of the minimum deviation and the maximum value of the deviation or the value of the standard deviation as the predicted improvement value, the possibility of raising the average of the objective variable can be evaluated.
　標準偏差の値の計算方法は、従来の方法を用いて行う。例えば、以下の式（３）を用いて標準偏差の値を計算する。 The standard deviation value is calculated using a conventional method. For example, the standard deviation value is calculated using the following equation (3).
Figure JPOXMLDOC01-appb-M000003
 　その後、改善値計算部１８０１は、計算した予測改善値を表示部１０９に送信する。
Figure JPOXMLDOC01-appb-M000003
 Thereafter, the improvement value calculation unit 1801 transmits the calculated predicted improvement value to the display unit 109.
　表示部１０９は、受信した予測改善値を顧客データセンタ１０のシステム端末１０２の表示装置に表示するための変換を行い、入出力部１０４を介してシステム端末１０２に送信する。 The display unit 109 performs conversion for displaying the received predicted improvement value on the display device of the system terminal 102 of the customer data center 10, and transmits it to the system terminal 102 via the input / output unit 104.
　図９は予測改善値を表示する画面の例を示した図である。図６と同じ部分には同じ符号を付してあり、表示する内容及び表示のための各構成の動作が同じであるので、説明を省略する。 FIG. 9 is a diagram showing an example of a screen displaying the predicted improvement value. The same parts as those in FIG. 6 are denoted by the same reference numerals, and the content to be displayed and the operation of each component for display are the same.
　目的変数６１と振れ幅閾値６２が設定され、計算実行ボタン６３が押されると、データ分析部１０８は、予測改善値の計算を行い、その結果を数値表示エリア９１に表示する。 When the objective variable 61 and the fluctuation threshold 62 are set and the calculation execution button 63 is pressed, the data analysis unit 108 calculates the predicted improvement value and displays the result in the numerical value display area 91.
　数値表示エリア９１は、予測改善値を表示するエリアであり、改善値計算部１８０１により計算された予測改善値を表示する。ここでは、計算した予測改善値をそのまま表示してもよいし、四捨五入して表示しても構わない。また、目的変数が生産性等の場合には、例えば、「１５パーセント減少／月」等の表示となる。必要に応じて目的変数の名称や指標ＩＤ等の目的変数を特定する情報を合わせて表示してもよい。この数値表示エリア９１により、データ解析システム１０５の導入効果を確認することが容易となる。 The numerical value display area 91 is an area for displaying the predicted improvement value, and displays the predicted improvement value calculated by the improvement value calculation unit 1801. Here, the calculated prediction improvement value may be displayed as it is, or may be displayed after rounding off. When the objective variable is productivity or the like, for example, “15% decrease / month” is displayed. Information specifying the objective variable such as the objective variable name and index ID may be displayed together as necessary. This numerical display area 91 makes it easy to confirm the effect of introducing the data analysis system 105.
　このように、本変形例に係る評価システム１０６は、時系列データについて所定の期間における偏差の値または中央値からの差を求める統計値計算部３０１と、偏差の値または中央値からの差に基づいてデータ解析システム１０５の導入による目的変数の改善を示す値である予測改善値を求める予測改善値計算部１８０１と、予測改善値を表示装置に表示する表示部１０９を有することを特徴としている。 As described above, the evaluation system 106 according to this modification includes the statistical value calculation unit 301 for obtaining a difference from the deviation value or the median value in a predetermined period for the time series data, and the difference from the deviation value or the median value. A prediction improvement value calculation unit 1801 that obtains a prediction improvement value that is a value indicating improvement of an objective variable based on the introduction of the data analysis system 105, and a display unit 109 that displays the prediction improvement value on a display device. .
　係る構成により、実際の目的変数の予想増加分の計算をすることが容易となる。これにより、顧客サイドは、導入費用と比較することが可能となるため、データ解析システムの導入の検討をすることが容易になる。 This configuration makes it easy to calculate the expected increase in actual objective variables. This makes it possible for the customer side to make a comparison with the introduction cost, making it easier to consider the introduction of the data analysis system.
　＜導入効果の費用変換に関する変形例＞
顧客が指定した目的変数が生産性の場合には、予測改善値は「１３」パーセント上昇等の割合で算出される。この上昇によりどの程度、職場のコストが削減されるかを事前に顧客に提示することは、顧客側のデータ解析システム導入の費用との比較をより容易にする。 <Variation regarding cost conversion of introduction effect>
When the objective variable designated by the customer is productivity, the predicted improvement value is calculated at a rate such as “13” percent increase. Providing customers in advance how much the cost of the workplace will be reduced by this increase makes it easier to compare with the cost of introducing a data analysis system on the customer side.
　ここでは、目的変数の属性が生産性の場合に、計算された予測改善値を予測改善値の分増加または減少した場合に削減されるコストである削減コストを計算する変形例について説明する。目的変数の属性が金額の場合には、上述の計算により求めた予測改善値をそのまま数値表示エリア９１に表示させるが、目的変数の属性が生産性等の金額以外の場合には、削減コストを求めて数値表示エリア９１に表示させる。なお、基本的なシステム構成は図１８と同じであるが、以下の点が異なる。 Here, a modified example of calculating a reduction cost, which is a cost to be reduced when the calculated prediction improvement value is increased or decreased by the prediction improvement value when the objective variable attribute is productivity, will be described. When the objective variable attribute is a monetary amount, the predicted improvement value obtained by the above calculation is displayed as it is in the numerical value display area 91. However, when the objective variable attribute is other than the monetary amount such as productivity, the reduction cost is reduced. It is obtained and displayed in the numerical value display area 91. The basic system configuration is the same as in FIG. 18, except for the following points.
　受付部１０７は、指定された目的変数の属性を受け付ける。顧客データセンタ１０内のシステム端末１０２で目的変数を指定する際に目的変数の属性を指定し、入出力部１０４を介して受付部１０７に送信される。また、目的変数の属性は、データ解析サービスセンタ１１内にて指定され、受付部１０７に送信されてもよい。こうすることで、顧客側の負担を減らすことが可能となる。 The accepting unit 107 accepts the attribute of the designated objective variable. When the objective variable is designated at the system terminal 102 in the customer data center 10, the objective variable attribute is designated and transmitted to the accepting unit 107 via the input / output unit 104. The attribute of the objective variable may be specified in the data analysis service center 11 and transmitted to the reception unit 107. This makes it possible to reduce the burden on the customer side.
　目的変数の属性とは、目的変数が金額に関するものか否かを表したものであり、大きく「金額」と「それ以外」に分けられる。受付部１０７は、受け付けた目的変数の属性を改善値計算部１８０１に送信する。 The objective variable attribute indicates whether or not the objective variable is related to a monetary amount, and is broadly divided into “amount” and “other”. The reception unit 107 transmits the received attribute of the objective variable to the improvement value calculation unit 1801.
　改善値計算部１８０１は受付部１０７から目的変数の属性を受け付け、削減コストを計算する。 The improvement value calculation unit 1801 receives the attribute of the objective variable from the reception unit 107 and calculates the reduction cost.
　改善値計算部１８０１は、コスト算出指標を予め保持している。コスト算出指標とは、目的変数の属性が金額に関するもの以外の場合に関して、その増加または減少によりどの程度費用が削減されるかを表した指標である。例えば、作業員一人あたりの単位時間あたりの作業個数が目的変数であった場合には、その作業個数の増加により減らすことが可能な作業員の作業時間が算出されるため、作業員の人件費と作業時間を掛け合わせることにより、削減されるコストが求められる。このように単位時間あたりの作業個数についての人件費の削減額がコスト算出指標の例である。コスト算出指標の他の例として、店舗入場者数１人あたりの平均購買額、機械停止確率１パーセントあたりの復帰作業費用等が挙げられる。 The improvement value calculation unit 1801 holds a cost calculation index in advance. The cost calculation index is an index that represents how much the cost is reduced by the increase or decrease in the case where the attribute of the objective variable is other than the monetary value. For example, if the number of work per unit time per worker is the objective variable, the work time of the worker that can be reduced by increasing the number of work is calculated. The cost to be reduced is calculated by multiplying the operation time. Thus, the amount of labor cost reduction for the number of work per unit time is an example of the cost calculation index. Other examples of the cost calculation index include an average purchase amount per store visitor, a return work cost per machine stop probability of 1%, and the like.
　コスト算出指標は顧客データセンタ１０側が保持している場合もあるため、その場合には、受付部１０７が入出力部１０４経由でコスト算出指標を受け付け、改善値計算部１８０１に送信する。 Since the customer data center 10 side may hold the cost calculation index, in this case, the receiving unit 107 receives the cost calculation index via the input / output unit 104 and transmits it to the improved value calculation unit 1801.
　次に、改善値計算部１８０１は、計算した予測改善値と、コスト算出指標とに基づき削減コストを計算する。例えば、予測改善値にコスト算出指標を掛け合わせることにより、目的変数の属性が金額に関するものでない場合の予測改善値を金額にて表現することができる。その後、計算した削減コストを表示部１０９に送信する。 Next, the improvement value calculation unit 1801 calculates a reduction cost based on the calculated predicted improvement value and the cost calculation index. For example, by multiplying the predicted improvement value by the cost calculation index, the predicted improvement value when the attribute of the objective variable is not related to the amount can be expressed by the amount. Thereafter, the calculated reduction cost is transmitted to the display unit 109.
　図１０は改善値計算部１８０１の削減コストの計算処理のフローを示した図の例である。 FIG. 10 is an example of a diagram illustrating a flow of a reduction cost calculation process of the improvement value calculation unit 1801.
　ステップ１００１では、改善値計算部１８０１により、統計値計算部３０１から偏差の値を受け付け、予測改善値を計算する。具体的な処理は既に述べているので省略する。 In step 1001, the improvement value calculation unit 1801 receives a deviation value from the statistical value calculation unit 301 and calculates a predicted improvement value. Since specific processing has already been described, it will be omitted.
　ステップ１００２では、改善値計算部１８０１により、受付部１０７から目的変数の属性を受け付ける。 In step 1002, the improvement value calculation unit 1801 receives the attribute of the objective variable from the reception unit 107.
　ステップ１００３では、改善値計算部１８０１により、ステップ１００２で受け付けた目的変数の属性が金額に関するものであるか否かの判断を行う。ここで、目的変数の属性が金額に関するものと判断された場合にはステップ１００４に移行し、金額以外に関するものと判断された場合にはステップ１００５に移行する。 In step 1003, the improvement value calculation unit 1801 determines whether or not the attribute of the objective variable received in step 1002 relates to the amount. If it is determined that the attribute of the objective variable is related to the amount, the process proceeds to step 1004. If it is determined that the attribute of the objective variable is related to other than the amount, the process proceeds to step 1005.
　ステップ１００４では、改善値計算部１８０１により、ステップ１００１にて計算した予測改善値を表示部１０９に送信する。 In step 1004, the improvement value calculation unit 1801 transmits the predicted improvement value calculated in step 1001 to the display unit 109.
　ステップ１００５では、改善値計算部１８０１により、ステップ１００１にて計算した予測改善値とコスト算出指標をもとに、削減コストを計算する。 In step 1005, the improvement value calculation unit 1801 calculates a reduction cost based on the predicted improvement value and the cost calculation index calculated in step 1001.
　ステップ１００６では、改善値計算部１８０１により、ステップ１００５にて計算した削減コストを表示部１０９に送信する。 In step 1006, the improvement value calculation unit 1801 transmits the reduction cost calculated in step 1005 to the display unit 109.
　このように、本変形例に係る評価システム１０６は、目的変数の属性を受け付ける受付部１０７と、目的変数の属性が金額に関するものでない場合には、時系列データが予測改善値の分増加または減少した場合に削減されるコストである削減コストを求める改善値計算部１８０１、削減コストを表示する表示部１０９を備えている。 As described above, the evaluation system 106 according to the present modification includes the receiving unit 107 that receives the attribute of the objective variable, and when the attribute of the objective variable is not related to the amount of money, the time series data increases or decreases by the predicted improvement value. In this case, an improvement value calculation unit 1801 for obtaining a reduction cost, which is a cost to be reduced, and a display unit 109 for displaying the reduction cost are provided.
　係る構成により、目的変数の属性が金額に関するものでない場合でも、顧客は解析システムの提示する施策による目的変数に関しての金額面での効果を事前に知ることができる。これにより、データ解析システムの導入費用との費用対効果を判断することがより容易となり、データ解析システムのスムースな導入に繋がる。 With such a configuration, even when the objective variable attribute is not related to the monetary amount, the customer can know in advance the monetary effect related to the objective variable by the measure presented by the analysis system. Thereby, it becomes easier to determine the cost-effectiveness with the introduction cost of the data analysis system, which leads to a smooth introduction of the data analysis system.
　データ解析システムによる解析は、状況変化への追従性を考慮して、解析期間幅を適切に設定してデータ解析をする必要があった。そこで、本実施例では、本発明の評価システムの別の例を示す。 In the analysis by the data analysis system, it was necessary to analyze the data by appropriately setting the analysis period width in consideration of the followability to the situation change. Therefore, in this embodiment, another example of the evaluation system of the present invention is shown.
　実施例１では、データ分析部によりデータ解析システムの導入効果を評価し、回帰式を作成していたが、実施例２では導入効果が大きいと判定された場合に、業務データをデータ解析システムが解析する解析期間幅を決定するための評価システムの内容について説明する。基本的なシステム構成は図３と同様であるが、以下の点で相違する。 In the first embodiment, the data analysis system evaluates the introduction effect of the data analysis system and creates the regression equation. However, in the second embodiment, when the introduction effect is determined to be large, the data analysis system converts the business data to the business data. The contents of the evaluation system for determining the analysis period width to be analyzed will be described. The basic system configuration is the same as that shown in FIG. 3 except for the following points.
　図１１は、本実施例に係る評価システム１０６の構成図の例である。評価システム１０６は、業務データの解析期間幅を決定し、決定した解析期間幅をデータ解析サービスセンタ１１内のデータ解析システム１０５に送信する。評価システム１０６は、図３の構成に加えて、期間候補作成部１１０１および期間決定部１１０２を更に備えている。 FIG. 11 is an example of a configuration diagram of the evaluation system 106 according to the present embodiment. The evaluation system 106 determines the analysis period width of the business data, and transmits the determined analysis period width to the data analysis system 105 in the data analysis service center 11. The evaluation system 106 further includes a period candidate creation unit 1101 and a period determination unit 1102 in addition to the configuration of FIG.
　期間候補作成部１１０１は、データ分析部１０８から導入効果の判定結果および目的変数に関する時系列データを受け付ける。そして、データ分析部１０８によって、導入効果の判定結果が大きいと判定された場合には、データ解析システム１０５が解析を行う期間の候補である期間候補を作成し、期間決定部１１０２に送信する。 The period candidate creation unit 1101 receives the introduction effect determination result and the time-series data regarding the objective variable from the data analysis unit 108. If the data analysis unit 108 determines that the introduction effect determination result is large, the data analysis system 105 creates a period candidate that is a period candidate for analysis, and transmits the period candidate to the period determination unit 1102.
　期間決定部１１０２は、期間候補作成部１１０１から目的変数に関する時系列データおよび期間候補を受け付ける。また、期間候補内の目的変数に関する時系列データの形状を基に、受け付けた期間候補が解析に適しているかどうかの判定を行う。さらに、解析に適していると判定された場合には、該当する期間候補の解析期間幅を、データ解析システム１０５に送信する。 The period determination unit 1102 receives time-series data and period candidates related to the objective variable from the period candidate creation unit 1101. Further, based on the shape of the time series data regarding the objective variable in the period candidate, it is determined whether or not the accepted period candidate is suitable for analysis. Further, when it is determined that the analysis is suitable, the analysis period width of the corresponding period candidate is transmitted to the data analysis system 105.
　解析期間幅とは、データ解析システム１０５が業務データ１０１の解析を逐次行う際に、利用する業務データの時間長のことであり、期間候補とは、解析期間幅の決定の際に仮に作成される期間の候補のことである。例えば、データ解析システム１０５は、目的変数に影響の大きい説明変数を抽出したり、その目的変数を上昇させるための施策を提示したりするために、特に解析期間幅内の業務データ１０１を解析する。 The analysis period width is the time length of the business data to be used when the data analysis system 105 sequentially analyzes the business data 101. The period candidate is temporarily created when the analysis period width is determined. It is a candidate for the period. For example, the data analysis system 105 particularly analyzes the business data 101 within the analysis period width in order to extract an explanatory variable having a large influence on the objective variable or present a measure for increasing the objective variable. .
　図１２は、評価システム１０６が、所定の期間の目的変数に関する時系列データの振れ幅が振れ幅閾値より大きいと判定した場合、つまりデータ解析システムの導入効果が高いと判定した場合に、解析期間幅の決定を行う処理のフローの例を示した図である。ステップ１２０１の処理が期間候補作成部１１０１によって行われ、ステップ１２０２～１２０４までの処理が期間決定部１１０２によって行われる。 FIG. 12 shows the analysis period when the evaluation system 106 determines that the amplitude of the time-series data related to the objective variable for a predetermined period is larger than the amplitude threshold, that is, when the introduction effect of the data analysis system is high. It is the figure which showed the example of the flow of the process which determines a width | variety. The processing in step 1201 is performed by the period candidate creation unit 1101, and the processing from steps 1202 to 1204 is performed by the period determination unit 1102.
　ステップ１２０１では、期間候補作成部１１０１がデータ分析部１０８から目的変数に関する時系列データを受け付け、時系列データのうちランダムに選定した一定の期間を期間候補として作成する。その後、期間候補および目的変数に関する時系列データを期間決定部１１０２に送信する。 In step 1201, the period candidate creation unit 1101 receives time series data related to the objective variable from the data analysis unit 108, and creates a certain period randomly selected from the time series data as a period candidate. Thereafter, time-series data regarding the period candidate and the objective variable is transmitted to the period determining unit 1102.
　ステップ１２０２では、期間決定部１１０２が期間候補作成部１１０１から期間候補および目的変数に関する時系列データを受け付け、受け付けた目的変数に関する時系列データについて期間候補内に極大点及び極小点が含まれるかを判定する。極大点および極小点が含まれているかの判定は既存の方法を用いて行う。 In step 1202, the period determining unit 1102 receives time series data related to the period candidate and the objective variable from the period candidate creating unit 1101, and whether the maximum and minimum points are included in the period candidate for the received time series data related to the objective variable. judge. The existing method is used to determine whether the local maximum point and the local minimum point are included.
　図１３は、回帰式作成部３０２により導入効果が大きいと判定された目的変数の時系列データの例を示した図である。この時系列データ１３０１は、統計値計算部３０１によって作成されたものである。 FIG. 13 is a diagram showing an example of time-series data of objective variables determined to have a large introduction effect by the regression equation creation unit 302. The time series data 1301 is created by the statistical value calculation unit 301.
　ここで、極大点および極小点を少なくとも１つずつ含む期間候補が、データ解析システム１０５による解析に適している理由について説明する。図１３のｔ１１～ｔ１２間やｔ１２～ｔ１３間のように極大点および極小点が含まれていない期間においては、その期間だけの業務データを用いてデータ解析を行っても目的変数の値が大きい状態と小さい状態の区別をすることが難しい。その一方、ｔ１０～ｔ１１間のように極大点１３０２および極小点１３０３が含まれている期間は、目的変数の大きい状態と小さい状態の区別をすることが容易であるため、データ解析に適した期間であることに発明者は想到した。すなわち、データ解析システムは、ｔ１０～ｔ１１間の極小点付近の目的変数の小さい部分の値を極大点付近の値に上昇させるような施策を提示しやすい。そこで、ｔ１０～ｔ１１間のように極大点１３０２および極小点１３０３が含まれている期間を解析期間幅に適していると評価することにした。このような期間の業務データを解析することにより、目的変数の増加および減少を説明する説明変数の抽出をすることがより容易になり、データ解析の質が向上する。 Here, the reason why a period candidate including at least one maximum point and one minimum point is suitable for analysis by the data analysis system 105 will be described. In a period that does not include the maximum and minimum points, such as between t11 and t12 and between t12 and t13 in FIG. 13, the value of the objective variable is large even if data analysis is performed using business data for only that period. It is difficult to distinguish between a state and a small state. On the other hand, the period in which the maximum point 1302 and the minimum point 1303 are included, such as between t10 and t11, is easy to distinguish between a large state and a small state of the objective variable. The inventor came up with that. That is, the data analysis system can easily present a measure for increasing the value of the small portion of the objective variable near the minimum point between t10 and t11 to the value near the maximum point. Therefore, it was decided that the period including the maximum point 1302 and the minimum point 1303, such as between t10 and t11, was suitable for the analysis period width. By analyzing the business data during such a period, it becomes easier to extract explanatory variables that explain the increase and decrease of the objective variable, and the quality of the data analysis is improved.
　ステップ１２０２で、ｔ１１～ｔ１２およびｔ１２～ｔ１３のように、期間候補内に極大点および極小点が少なくとも１つずつ含まれていないと判定された場合には、ステップ１２０１に戻り、期間候補の作成から処理を繰り返す。ｔ１０～ｔ１１のように、期間候補内に極大点および極小点が少なくとも１つずつ含まれていると判定された場合には、ステップ１２０３に進む。 If it is determined in step 1202 that at least one local maximum point and local minimum point are not included in the period candidates, such as t11 to t12 and t12 to t13, the process returns to step 1201 to generate the period candidates. Repeat the process. When it is determined that at least one local maximum point and local minimum point are included in the period candidates as in t10 to t11, the process proceeds to step 1203.
　ステップ１２０３では、期間決定部１１０２はステップ１２０２で極大点および極小点が少なくとも１つずつ含まれると判定された期間候補を解析期間幅として決定する。 In step 1203, the period determination unit 1102 determines the period candidate determined in step 1202 as including at least one local maximum point and local minimum point as the analysis period width.
　ステップ１２０４では、期間決定部１１０２により決定された解析期間幅をデータ解析システム１０５に送信する。 In step 1204, the analysis period width determined by the period determination unit 1102 is transmitted to the data analysis system 105.
　このように、本実施例に係る評価システム１０６は、回帰式作成部３０２が、振れ幅が振れ幅閾値より大きいと判定した場合に、データ解析システム１０５が解析を行う期間の候補である期間候補を作成する期間候補作成部１１０１と、期間候補の期間内の時系列データについて、極大点および極小点が少なくとも１つずつ含まれている場合には、期間候補をデータ解析システム１０５が解析を行う解析期間幅に決定する期間決定部１１０２を備えている。 As described above, in the evaluation system 106 according to the present embodiment, when the regression equation creation unit 302 determines that the amplitude is larger than the amplitude threshold, the period candidate that is a candidate for the period in which the data analysis system 105 performs the analysis. If the period candidate creation unit 1101 that creates the period candidate and the time series data in the period candidate period include at least one local maximum point and local minimum point, the data analysis system 105 analyzes the period candidate. A period determining unit 1102 that determines the analysis period width is provided.
　係る構成により、データ解析システムの導入効果が大きいと評価した場合に、その導入効果を更に大きくするためのデータの解析期間幅を適切に決定することが可能となる。更に、データの解析期間幅を適切に決定することで、計算処理量の見積りが可能となり、コンピュータ資源等のリソースを有効に利用することが可能となる。 With such a configuration, when it is evaluated that the introduction effect of the data analysis system is great, it is possible to appropriately determine the data analysis period width for further increasing the introduction effect. Furthermore, by appropriately determining the data analysis period width, it is possible to estimate the amount of calculation processing, and it is possible to effectively use resources such as computer resources.
　実施例１では、データ解析システムの導入の効果の判定をするために、導入効果を判定し、回帰式を作成する評価システムについて記載してきた。次に、実際に顧客の業務データを用いて解析を行うデータ解析システムの例を示す。 In Example 1, in order to determine the effect of introducing the data analysis system, an evaluation system that determines the effect of introduction and creates a regression equation has been described. Next, an example of a data analysis system that actually performs analysis using customer business data is shown.
　図１５は、データ解析システム１０５の詳細構成を含む本実施例の概要を示した図である。顧客データセンタ１０および外部のデータ解析システム１０５のそれぞれの構成およびデータ処理について説明する。 FIG. 15 is a diagram showing an outline of the present embodiment including the detailed configuration of the data analysis system 105. The configuration and data processing of the customer data center 10 and the external data analysis system 105 will be described.
　顧客データセンタ１０には、業務データ１０１が格納されている。業務データ１０１は、過去データ１０１Ａと当日データ１０１Ｂに分かれて格納される。区別して格納されていない場合には、後に説明する取得部１２０Ａによって、業務データ１０１を受け付けた後に、過去データ１０１Ａおよび当日データ１０１Ｂに区別する。また、顧客の業務システム１１０は、顧客側の従業員に業務の指示を与える業務指示部１１０Ａおよび業務の指示に対しての業績結果を確認する業績管理部１１０Ｂにより構成される。これらの内容はシステム端末１０２より確認できる。なお、図１５は本実施例の顧客データセンタ１０およびデータ解析システム１０５のそれぞれの構成の一例であり、データ解析システム１０５が顧客データセンタ１０内の業務システム１１０内に組み込まれていても構わない。このような構成とすることで、顧客の業務データのやり取りを顧客データセンタ内で完結させることが可能となる。 The business data 101 is stored in the customer data center 10. The business data 101 is stored separately as past data 101A and current day data 101B. If they are not stored separately, the acquisition unit 120A, which will be described later, distinguishes the past data 101A and the current day data 101B after receiving the business data 101. The customer business system 110 includes a business instruction unit 110A that gives business instructions to employees on the customer side, and a performance management unit 110B that confirms the business results for the business instructions. These contents can be confirmed from the system terminal 102. FIG. 15 shows an example of the configuration of each of the customer data center 10 and the data analysis system 105 of this embodiment, and the data analysis system 105 may be incorporated in the business system 110 in the customer data center 10. . With this configuration, it is possible to complete the exchange of customer business data within the customer data center.
　データ解析システム１０５は、案件管理部１２０、モデル生成部１３０および変数最適化部１４０を備えている。 The data analysis system 105 includes a case management unit 120, a model generation unit 130, and a variable optimization unit 140.
　案件管理部１２０においては、必要なデータを用いて生産性または売上等の目的変数の最適化処理を業務毎に行う。案件管理部１２０は、取得部１２０Ａにより顧客データセンタ１０より過去データ１０１Ａおよびどの変数を目的変数とするかの指定を取得して、モデル生成部１３０に送信する。 The project management unit 120 performs optimization processing of objective variables such as productivity or sales for each business using necessary data. The case management unit 120 acquires the specification of the past data 101A and which variable is the target variable from the customer data center 10 by the acquisition unit 120A, and transmits it to the model generation unit 130.
　モデル生成部１３０は、過去データ取得部１３１、複合説明変数生成部１３２、変数生成ロジックデータベース１３３、複合説明変数データベース１３４、統計モデリング部１３５およびモデルデータベース１３６を備えている。 The model generation unit 130 includes a past data acquisition unit 131, a composite explanatory variable generation unit 132, a variable generation logic database 133, a composite explanatory variable database 134, a statistical modeling unit 135, and a model database 136.
　モデル生成部１３０においては、過去データ取得部１３１にて、過去データ１０１Ａ及び目的変数の指定を受信し、複合説明変数生成部１３２において受信した過去データ１０１Ａを複合的に組み合わせ、複合説明変数に変換した後に複合説明変数データベース１３４に格納する。複合説明変数の生成には格納されている変数生成ロジック１３３が使用される。変数生成ロジック１３３は、予め設定されたロジックを使用することも可能であるが、業績管理部１１０Ｂにおいて管理する業績結果と連動することによりダイナミックに変更したロジックを使用することもできる。ここでは、例えば、過去データ１０１Ａについて、データ同士を複合的に組み合わせて、目的変数に影響を与え得る新たな複合説明変数を生成する処理等が行われる。例えば、１０００個から１００万個を超える大量の複合説明変数を自動生成する。 In the model generation unit 130, the past data acquisition unit 131 receives the specification of the past data 101A and the objective variable, and combines the past data 101A received by the composite explanatory variable generation unit 132 in a composite manner and converts it into a composite explanatory variable. After that, it is stored in the composite explanatory variable database 134. A stored variable generation logic 133 is used to generate the composite explanatory variable. The variable generation logic 133 can use a preset logic, but can also use a logic that is dynamically changed by interlocking with a result of achievement managed in the achievement management unit 110B. Here, for example, for the past data 101A, a process of generating a new composite explanatory variable that can affect the target variable by combining the data in a composite manner is performed. For example, a large number of complex explanatory variables exceeding 1000 to 1 million are automatically generated.
　複合説明変数の生成方法としては、特許文献１に記載のような、条件オペレータ、対象オペレータおよび演算オペレータの３つのオペレータのセットによる説明変数生成処理を行っても良いし、既存の他の方法で複合説明変数を生成しても良い。複合説明変数の例として、室温が１５度以下であるか等のある説明変数についての条件付きの変数や、室温が１５度以下であり、かつ、作業個数が１００個毎分より大きい場合には、目的変数が大きい傾向があるといった場合の「室温×作業個数」を示す変数等が挙げられる。 As a method for generating a composite explanatory variable, an explanatory variable generation process by a set of three operators, a conditional operator, a target operator, and a calculation operator, as described in Patent Document 1, may be performed, or other existing methods may be used. A composite explanatory variable may be generated. Examples of complex explanatory variables include conditional variables for certain explanatory variables, such as whether the room temperature is 15 degrees or less, or when the room temperature is 15 degrees or less and the number of operations is greater than 100 per minute. And a variable indicating “room temperature × number of work” when the objective variable tends to be large.
　更に、統計モデリング部１３５においては、複合説明変数データベース１３４に格納されている複合説明変数の中から、顧客の指定した売上や生産性等に関する変数（目的変数ｙ）に影響する変数を統計的に選択し、それらの関係性をモデルとして作成し、作成したモデルをモデルデータベース１３６に格納する。ここでのモデルは、目的変数と複合説明変数との関係を表せるモデルであれば良く、例えば、回帰式等が挙げられる。 Further, the statistical modeling unit 135 statistically analyzes a variable affecting the variable (objective variable y) related to sales, productivity, etc. designated by the customer from the composite explanatory variables stored in the composite explanatory variable database 134. These are selected and their relationship is created as a model, and the created model is stored in the model database 136. The model here may be a model that can represent the relationship between the objective variable and the composite explanatory variable, and examples thereof include a regression equation.
　業務判断において考慮すべき要因が多数あると、膨大な選択肢からの判断が必要である。例えば、倉庫に１００件の出荷依頼がある場合には、１０の７９乗個（１００の階乗個）という膨大な作業順序の組合せの中から、作業効率が良い作業順序を選択する必要がある。このモデルを利用すると、これらの選択肢の優劣を判断することが可能となる。 If there are many factors that should be taken into consideration in the business judgment, it is necessary to judge from a huge number of options. For example, when there are 100 shipment requests in the warehouse, it is necessary to select a work order with good work efficiency from among a huge work order combination of 10 79 (100 factorial). . By using this model, it is possible to judge the superiority or inferiority of these options.
　また、データ解析システム１０５がウェブＡＰＩを介して業務システム１１０からデータと設定情報を受信することにより、モデルを作成する構成にしても構わない。このような構成にすることで、業務ごとに複雑な個別プログラムを有したデータ解析システム１０５を作成する必要が無くなることで最適化問題の専門知識が無くてもデータ解析を行うことが可能となり、ソフトウェア開発の労力を抑えることが可能となる。 Further, the data analysis system 105 may be configured to create a model by receiving data and setting information from the business system 110 via the web API. By adopting such a configuration, it becomes unnecessary to create the data analysis system 105 having complicated individual programs for each business, so that it is possible to perform data analysis even without specialized knowledge of optimization problems. Software development effort can be reduced.
　案件管理部１２０は、モデル生成部１３０において生成されたモデルを取得して、変数最適化部１４０に送信する。次に変数最適化部１４０における処理に関して説明する。 The matter management unit 120 acquires the model generated by the model generation unit 130 and transmits the model to the variable optimization unit 140. Next, processing in the variable optimization unit 140 will be described.
　変数最適化部１４０は、当日データ取得部１４１、モデル取得部１４２、組合せ変更・複合説明変数生成部１４３および目的変数評価部１４４を備えている。 The variable optimization unit 140 includes a day data acquisition unit 141, a model acquisition unit 142, a combination change / composite explanatory variable generation unit 143, and an objective variable evaluation unit 144.
　当日データ取得部１４１は、顧客データセンタ１０から当日データ１０１Ｂを受け付ける。また、モデル取得部１４２は、案件管理部１２０からモデル１３６を受け付ける。顧客データセンタ１０の業務データ１０１が過去データ１０１Ａと当日データ１０１Ｂで分けて格納されていない場合には、案件管理部１２０から区別された当日データを当日データ取得部１４１が受け付ける。 The day data acquisition unit 141 receives the day data 101B from the customer data center 10. Further, the model acquisition unit 142 receives the model 136 from the case management unit 120. When the business data 101 of the customer data center 10 is not stored separately for the past data 101A and the current day data 101B, the current day data acquisition unit 141 receives the current day data distinguished from the case management unit 120.
　このように取得されたモデル１３６および当日データ１０１Ｂは、組合せ変更・複合説明変数生成部１４３に送信される。 The model 136 and the day data 101B acquired in this way are transmitted to the combination change / composite explanatory variable generation unit 143.
　組合せ変更・複合説明変数生成部１４３は、業務タスクの組合せパターンを変更し、実際の当日データを入力して、組合せパターンの変更に該当する複合説明変数を計算する。ここでは、過去データを用いて作成したモデルに対して、当日のデータを入力して複合説明変数を計算しているため、当日の目的変数に対する影響を考慮したモデルの作成が出来ると同時に、業務タスクの組合せのパターンの変更に該当する複合説明変数を計算しているため、後で説明する目的変数の評価および業務タスクの組合せのパターンの変更の繰り返し処理により、目的変数が改善する組合せのパターンを抽出することが可能となる。 The combination change / composite explanatory variable generation unit 143 changes a combination pattern of business tasks, inputs actual day data, and calculates a composite explanatory variable corresponding to the change of the combination pattern. Here, since the compound explanatory variables are calculated by inputting the data for the current day against the model created using the past data, it is possible to create a model that takes into account the influence on the objective variable for the day, and at the same time Since the complex explanatory variable corresponding to the change of task combination pattern is calculated, the combination pattern that the objective variable is improved by repeating the evaluation of the objective variable and the change of the business task combination pattern described later Can be extracted.
　次に、目的変数評価部１４４は、業務タスクの組合せに対応する目的変数について、組合せ変更・複合説明変数生成部１４３にて計算された複合説明変数を用いて評価する。目的変数評価部１４４は、業務タスクの組合せに対応する目的変数が改善しているかどうかを判断し、目的変数の値が改善していると判断した場合には、その業務タスクの組合せを業務システム１１０に送信する。目的変数の値が改善していないと判断した場合には、業務タスクの組合せの変更を行わない。ここでは、業務タスクの変更を１回しか行っていないが、これを繰り返し行う処理も可能である。繰り返し処理を行う目的変数評価部１４４の処理の例については、１４５～１４７に示している。ここでは、業務タスクの組合せパターンは逐次変更され、目的変数が改善されるように探索的な処理が繰り返し実施される。具体的な処理としては、組合せ変更・複合説明変数生成部１４３において、業務タスクの組合せが変更され、複合説明変数を計算し、目的変数が改善されたかどうかのチェックが行われ（１４６）、目的変数が改善されない場合には、組合せの変更を行わず前回組合せが保持される（１４７）。 Next, the objective variable evaluation unit 144 evaluates the objective variable corresponding to the combination of business tasks by using the composite explanatory variable calculated by the combination change / composite explanatory variable generation unit 143. The objective variable evaluation unit 144 determines whether or not the objective variable corresponding to the combination of business tasks has been improved. If the objective variable evaluation unit 144 determines that the value of the objective variable has been improved, 110. If it is determined that the value of the objective variable has not improved, the business task combination is not changed. In this example, the business task is changed only once, but it is also possible to perform this process repeatedly. Examples of processing of the objective variable evaluation unit 144 that performs repetitive processing are shown in 145 to 147. Here, the combination pattern of business tasks is sequentially changed, and exploratory processing is repeatedly performed so that the objective variable is improved. Specifically, in the combination change / composite explanatory variable generation unit 143, a combination of business tasks is changed, a composite explanatory variable is calculated, and whether the objective variable is improved is checked (146). If the variable is not improved, the previous combination is held without changing the combination (147).
　前回組み合わせに戻す処理（１４７）において、目的変数が改善されていない場合にも、ある確率で組合せの変更を行うことで局所最適に陥らないようにすることも可能である。これらの繰り返し処理は、目的変数が収束するか、または繰り返しの上限回数に達するかどうか、という終了条件によって判定され（１４５）、終了となった場合には、その組合せを最適化の結果として、業務システムに１１０に送信する。終了条件としては、例えば、目的変数の改善レートが所定の改善閾値を超えた場合等が挙げられる。 In the process of returning to the previous combination (147), even when the objective variable has not been improved, it is possible to prevent the local optimum from falling by changing the combination with a certain probability. These iterative processes are determined by an end condition whether the objective variable converges or the upper limit number of iterations is reached (145). When the end is reached, the combination is determined as an optimization result. Transmit to 110 to the business system. Examples of the termination condition include a case where the improvement rate of the objective variable exceeds a predetermined improvement threshold.
　図１６に、以上の探索処理を行った際の、目的変数の改善レートを図示しており、探索処理開始時からどの程度目的変数が改善したかを縦軸に示してある。図１６における目的変数は、例えば作業時間等の、値の小さい方が好ましい変数である。ここでは、改善閾値１６０１として、閾値５パーセントを設定した例を示している。この改善閾値１６０１を下回ることを終了条件として設定しており、図１６の例では、約１８０００回の組合せ変更の試行により有為な目的変数の改善が行われていると判断することができる。 FIG. 16 shows the improvement rate of the objective variable when the above search process is performed, and the vertical axis indicates how much the objective variable has improved since the start of the search process. The objective variable in FIG. 16 is a variable with a smaller value such as work time, for example. Here, an example in which a threshold of 5% is set as the improvement threshold 1601 is shown. It is set as an end condition that the value is less than the improvement threshold 1601. In the example of FIG. 16, it can be determined that a significant objective variable has been improved by about 18,000 combinations of changes.
　このように繰り返し処理を行うことで、目的変数が最適化された業務のタスクの組合せパターンを抽出することが可能となる。 繰 り 返 し By repeating the process in this way, it is possible to extract task task combination patterns with optimized objective variables.
　図１７には、上記の探索処理で改善される業務の一例として、物流倉庫においてのピッキング処理順序（カートＮｏ）を最適化する処理を示している。ピッキング処理のカートＮｏを入れ替えて、その後にモデル１３６を用いて全体作業時間の評価をし、入れ替えを保持する又は入れ替え前の組合せに戻す処理を繰り返すことにより、作業時間が最適化されるようなピッキング処理順序を得ることができる。なお、この例においては、目的変数としての全体作業時間は、ピッキング作業エリアにおけるカートの各棚への滞在時間（説明変数）の組合せ、として表現される場合を示している。ここでの目的変数と説明変数の組合せは一例であり、売上等の目的変数や店員の滞在エリアへの滞在時間等の説明変数の組合せであっても構わない。 FIG. 17 shows a process for optimizing the picking process order (cart No.) in the distribution warehouse as an example of the business improved by the above search process. The work time is optimized by replacing the picking cart No. and then evaluating the overall work time using the model 136 and repeating the process of holding the replacement or returning to the combination before the replacement. Picking processing order can be obtained. In this example, the total work time as the objective variable is expressed as a combination of the stay time (explanatory variable) on each shelf of the cart in the picking work area. The combination of the objective variable and the explanatory variable here is an example, and may be a combination of the objective variable such as sales and the explanatory variable such as the staying time of the clerk in the staying area.
　このように、本実施例に係るデータ解析システム１０５は、業務データおよび業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける取得部１２０Ａと、業務データのうちの過去のデータを用いて、目的変数と業務データ内の複数の変数との関係性をモデルとして作成する統計モデリング部１３５と、モデルおよび業務データのうちの当日のデータに基づいて、業務のタスクのパターンを変更した場合の目的変数の変化を計算し、目的変数が改善した場合に、パターンの変更を維持する目的変数評価部１４４を備えている。 As described above, the data analysis system 105 according to this embodiment includes the acquisition unit 120A that receives designation of which variable is the target variable among the business data and the plurality of variables in the business data, and the past of the business data. Statistical modeling unit 135 that creates the relationship between the objective variable and multiple variables in the business data as a model, and the task pattern of the business based on the data of the day of the model and the business data The objective variable evaluation unit 144 is provided that calculates the change of the objective variable when the variable is changed, and maintains the pattern change when the objective variable is improved.
　係る構成により、顧客側で実際に業務のタスクのパターンを変えて効果を判断し、試行錯誤的に最善の業務のタスクのパターンを判断することなく、自動でより目的変数に効果の大きい業務のタスクのパターンを算出することが可能となる。 With such a configuration, the customer actually changes the task pattern of the business to determine the effect, and without automatically determining the task pattern of the best business on a trial and error basis, It is possible to calculate a task pattern.
　なお、実施例１、実施例２で説明したように、データ解析システム１０５においては、過去データ１０１Aの中から評価システム１０６にて決定された必要な変数のみを扱うことができ、また、評価システム１０６にて決定された規定された解析期間幅を使用することができる。従って、事前に必要な計算リソースを見積もることができ、コンピュータ資源等を有効に利用することが可能となる。 As described in the first and second embodiments, the data analysis system 105 can handle only necessary variables determined by the evaluation system 106 from the past data 101A. A defined analysis period width determined at 106 can be used. Therefore, necessary calculation resources can be estimated in advance, and computer resources and the like can be used effectively.
　１０　顧客データセンタ、１１　データ解析サービスセンタ
　１０１　業務データ、１０２　システム端末、
　１０５　データ解析システム、１０６　評価システム、１０７　受付部、
　１０８　データ分析部、１０９　表示部、３０１　統計値計算部、
　３０２　回帰式作成部、４０１　目的変数ａの時系列データ、
　５０１　業務データ・目的変数受付ステップ、
　５０２　統計値計算ステップ、５０３～５０５　効果判定ステップ、
　５０６　回帰式作成ステップ、５０７　判定結果送信ステップ、
　５０８　判定結果及び回帰式送信ステップ、
　５０９　表示ステップ、６０１　目的変数指定エリア、
　６０２　導入効果表示エリア、６４　結果表示エリア、
　６７　回帰式表示エリア、９１　数値表示エリア、
　１００２　属性受付ステップ、１００５　削減コスト計算ステップ、
　１１０１　期間候補作成部、１１０２　期間決定部、
　１２０１　期間候補作成ステップ、
　１２０２、１２０３　解析期間幅決定ステップ、
　１４０１　ＰＯＳテーブル、１４０２　作業テーブル、
　１０１Ａ　過去データ、１０１Ｂ　当日データ、
　１１０Ａ　業務指示部、１１０Ｂ　業績管理部、１２０　案件管理部、
　１２０Ａ　取得部、１３０　モデル生成部、１３１　過去データ取得部、
　１３２　複合説明変数生成部、１３３　変数生成ロジックデータベース、
　１３４　複合説明変数データベース、１３５　統計モデリング部、
　１３６　モデルデータベース、１４０　変数最適化部、
　１４１　当日データ取得部、１４２　モデル取得部、
　１４３　組合せ変更・複合説明変数生成部、１４４　目的変数評価部、
　１４６　目的変数改善表示、１６０１　改善閾値、
　１８０１　改善値計算部 10 Customer Data Center, 11 Data Analysis Service Center 101 Business Data, 102 System Terminal,
105 data analysis system, 106 evaluation system, 107 reception unit,
108 data analysis unit, 109 display unit, 301 statistical value calculation unit,
302 regression equation creation unit 401 time series data of objective variable a,
501 Business data / objective variable reception step,
502 statistical value calculation step, 503 to 505 effect determination step,
506 regression equation creation step, 507 judgment result transmission step,
508 judgment result and regression transmission step,
509 display step, 601 objective variable designation area,
602 Introduction effect display area, 64 result display area,
67 Regression equation display area, 91 Numerical value display area,
1002 attribute reception step, 1005 reduction cost calculation step,
1101 period candidate creation unit, 1102 period determination unit,
1201 Period candidate creation step,
1202, 1203 Analysis period width determination step,
1401 POS table, 1402 work table,
101A historical data, 101B current day data,
110A business instruction department, 110B performance management department, 120 project management department,
120A acquisition unit, 130 model generation unit, 131 past data acquisition unit,
132 composite explanation variable generation part, 133 variable generation logic database,
134 complex explanatory variable database, 135 statistical modeling department,
136 model database, 140 variable optimization section,
141 data acquisition unit on the day, 142 model acquisition unit,
143 combination change / composite explanatory variable generation unit, 144 objective variable evaluation unit,
146 Objective variable improvement display, 1601 Improvement threshold,
1801 Improvement value calculator

Claims (15)

  1. 　データ解析システムの導入効果を表示する評価システムであって、
    　業務データおよび前記業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける受付部と、
    　前記目的変数に関する時系列データの所定の期間における振れ幅を求める統計値計算部と、
    　前記振れ幅が所定の振れ幅閾値より大きい場合に、前記所定の期間内の前記目的変数、および、前記目的変数と相関のある前記業務データ内の変数である説明変数を基に回帰式を作成する回帰式作成部と、
    　前記回帰式を表示装置に表示する表示部と、を有することを特徴とする評価システム。     An evaluation system that displays the effect of introducing a data analysis system,
    A reception unit for receiving designation of which variable is the target variable among the business data and a plurality of variables in the business data;
    A statistical value calculation unit for obtaining a fluctuation in a predetermined period of time-series data regarding the objective variable;
    When the amplitude is larger than a predetermined amplitude threshold, a regression equation is created based on the objective variable within the predetermined period and an explanatory variable that is a variable in the business data correlated with the objective variable A regression equation creation unit,
    And a display unit that displays the regression equation on a display device.
  2. 　請求項１に記載の評価システムであって、
    　前記統計値計算部は、前記時系列データについて前記所定の期間における分散値を求め、
    　前記回帰式作成部は、前記分散値が所定の分散閾値より大きい場合に、前記所定の期間内の前記回帰式を作成することを特徴とする評価システム。     The evaluation system according to claim 1,
    The statistical value calculation unit obtains a variance value in the predetermined period for the time series data,
    The evaluation system, wherein the regression equation creation unit creates the regression equation within the predetermined period when the variance value is greater than a predetermined variance threshold.
  3. 　請求項１に記載の評価システムであって、
    　前記受付部は、顧客から前記振れ幅閾値を受け付け、
    　前記回帰式作成部は、前記振れ幅が前記顧客から受け付けた前記振れ幅閾値より大きい場合に前記所定の期間内の前記回帰式を作成することを特徴とする評価システム。     The evaluation system according to claim 1,
    The reception unit receives the deflection threshold from a customer,
    The evaluation system according to claim 1, wherein the regression equation creation unit creates the regression equation within the predetermined period when the amplitude is larger than the amplitude threshold received from the customer.
  4. 　請求項１に記載の評価システムであって、
    　前記回帰式作成部が、前記振れ幅が前記振れ幅閾値より大きいと判定した場合に、前記データ解析システムが解析を行う期間の候補である期間候補を作成する期間候補作成部と、
    　前記期間候補の期間内の前記時系列データについて、極大点および極小点が少なくとも１つずつ含まれている場合には、前記期間候補を前記データ解析システムが解析を行う期間である解析期間幅に決定する期間決定部を更に有することを特徴とする評価システム。     The evaluation system according to claim 1,
    When the regression equation creation unit determines that the amplitude is greater than the amplitude threshold, a period candidate creation unit that creates a period candidate that is a candidate for a period in which the data analysis system performs analysis;
    When the time series data within the period candidate includes at least one local maximum point and local minimum point, the period candidate is set to an analysis period width that is a period during which the data analysis system analyzes the period candidate. An evaluation system further comprising a period determining unit for determining.
  5. 　請求項１に記載の評価システムであって、
    　前記回帰式作成部は、前記振れ幅が前記所定の振れ幅閾値より大きい場合に前記導入効果が大きいと評価し、
    　前記表示部は、前記導入効果を前記表示装置に表示することを特徴とする評価システム。     The evaluation system according to claim 1,
    The regression equation creation unit evaluates that the introduction effect is large when the amplitude is larger than the predetermined amplitude threshold,
    The display system displays the introduction effect on the display device.
  6. 　請求項１に記載の評価システムであって、
    　前記説明変数は、前記業務データ内の変数を組合せて生成した複合説明変数であることを特徴とする評価システム。     The evaluation system according to claim 1,
    The evaluation system is a composite explanatory variable generated by combining variables in the business data.
  7. 　データ解析システムの導入効果を評価する評価方法であって、
    　業務データおよび前記業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける受付ステップと、
    　前記目的変数に関する時系列データの所定の期間における振れ幅を求める統計値計算ステップと、
    　前記振れ幅が所定の振れ幅閾値より大きい場合に、前記所定の期間内の前記目的変数、および、前記目的変数と相関のある前記業務データ内の変数である説明変数を基に回帰式を作成する回帰式作成ステップと、
    　前記回帰式を表示装置に表示する表示ステップと、を有することを特徴とする評価方法。     An evaluation method for evaluating the effect of introducing a data analysis system,
    A reception step of receiving designation of which variable is the target variable among the business data and a plurality of variables in the business data;
    A statistical value calculating step for obtaining a fluctuation width in a predetermined period of the time-series data regarding the objective variable;
    When the amplitude is larger than a predetermined amplitude threshold, a regression equation is created based on the objective variable within the predetermined period and an explanatory variable that is a variable in the business data correlated with the objective variable A regression equation creation step,
    A display step of displaying the regression equation on a display device.
  8. 　請求項７に記載の評価方法であって、
    　前記統計値計算ステップにおいて、前記時系列データについて前記所定の期間における分散値を求め、
    　前記回帰式作成ステップにおいて、前記分散値が所定の分散閾値より大きい場合に、前記所定の期間内の前記回帰式を作成することを特徴とする評価方法。     The evaluation method according to claim 7,
    In the statistical value calculating step, a variance value in the predetermined period is obtained for the time series data,
    In the regression equation creating step, the regression equation within the predetermined period is created when the variance value is larger than a predetermined variance threshold value.
  9. 　請求項７に記載の評価方法であって、
    　前記受付ステップにおいて、顧客から前記振れ幅閾値を受け付け、
    　前記回帰式作成ステップにおいて、前記振れ幅が前記顧客から受け付けた前記振れ幅閾値より大きい場合に前記所定の期間内の前記回帰式を作成することを特徴とする評価方法。     The evaluation method according to claim 7,
    In the receiving step, the swing threshold is received from a customer;
    In the regression equation creating step, the regression equation within the predetermined period is created when the amplitude is larger than the amplitude threshold received from the customer.
  10. 　請求項７に記載の評価方法であって、
    　前記回帰式作成ステップにおいて、前記振れ幅が前記振れ幅閾値より大きいと判定した場合に、前記データ解析システムが解析を行う期間の候補である期間候補を作成する期間候補作成ステップと、
    　前記期間候補の期間内の前記時系列データについて、極大点および極小点が少なくとも１つずつ含まれている場合に、前記期間候補を前記データ解析システムが解析を行う期間である解析期間幅に決定する期間決定ステップを更に有することを特徴とする評価方法。     The evaluation method according to claim 7,
    In the regression equation creation step, when it is determined that the amplitude is larger than the amplitude threshold, a period candidate creation step of creating a period candidate that is a candidate for a period in which the data analysis system performs analysis;
    When the time series data within the period candidate includes at least one local maximum point and local minimum point, the period candidate is determined as an analysis period width in which the data analysis system performs analysis. The evaluation method further comprising a period determining step.
  11. 　請求項７に記載の評価方法であって、
    　前記回帰式作成ステップにおいて、前記振れ幅が前記所定の振れ幅閾値より大きい場合に前記導入効果が大きいと評価し、
    　前記表示ステップにおいて、前記導入効果を前記表示装置に表示することを特徴とする評価方法。     The evaluation method according to claim 7,
    In the regression equation creation step, when the runout is larger than the predetermined runout threshold, it is evaluated that the introduction effect is large,
    In the display step, the introduction effect is displayed on the display device.
  12. 　請求項７に記載の評価方法であって、
    　前記説明変数は、前記業務データ内の変数を組合せて生成した複合説明変数であることを特徴とする評価方法。     The evaluation method according to claim 7,
    The evaluation method is characterized in that the explanatory variable is a composite explanatory variable generated by combining variables in the business data.
  13. 　業務データおよび前記業務データ内の複数の変数のうちどの変数を目的変数とするかの指定を受け付ける取得部と、
    　前記業務データのうちの過去のデータを用いて、前記目的変数と前記業務データ内の複数の変数との関係性をモデルとして作成する統計モデリング部と、
    　前記モデルおよび前記業務データのうちの当日のデータに基づいて、業務のタスクのパターンを変更した場合の前記目的変数の変化を計算する目的変数評価部と、を有し、
    　前記目的変数評価部は、前記目的変数が改善した場合に、前記パターンの変更を維持することを特徴とするデータ解析システム。     An acquisition unit that receives designation of business data and which of the plurality of variables in the business data is a target variable;
    A statistical modeling unit that creates a relationship between the objective variable and a plurality of variables in the business data by using past data of the business data; and
    An objective variable evaluation unit that calculates a change in the objective variable when a task pattern of the task is changed based on data of the day of the model and the task data; and
    The objective variable evaluation unit maintains the pattern change when the objective variable is improved.
  14. 　請求項１３に記載のデータ解析システムであって、
    　前記目的変数評価部は、前記目的変数が所定の改善閾値を超えて収束するか、または、前記パターンの変更の回数が所定の上限を超えるまで、前記パターンの変更を繰り返すことを特徴とするデータ解析システム。     The data analysis system according to claim 13,
    The objective variable evaluation unit repeats the change of the pattern until the objective variable converges beyond a predetermined improvement threshold or the number of changes of the pattern exceeds a predetermined upper limit. Analysis system.
  15. 　請求項１３に記載のデータ解析システムであって、
    　前記業務データ内の複数の変数を組み合わせた複合説明変数を生成する複合説明変数生成部を更に有し、
    　前記統計モデリング部が前記モデルの作成に用いる前記業務データ内の複数の変数は、前記複合説明変数であることを特徴とするデータ解析システム。     The data analysis system according to claim 13,
    A composite explanatory variable generating unit that generates a composite explanatory variable combining a plurality of variables in the business data;
    A data analysis system, wherein a plurality of variables in the business data used by the statistical modeling unit to create the model are the composite explanatory variables.
PCT/JP2015/074319 2015-08-28 2015-08-28 Evaluation system, evaluation method, and data analysis system WO2017037768A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/JP2015/074319 WO2017037768A1 (en) 2015-08-28 2015-08-28 Evaluation system, evaluation method, and data analysis system
JP2016574300A JP6326510B2 (en) 2015-08-28 2015-08-28 Evaluation system, evaluation method, and data analysis system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2015/074319 WO2017037768A1 (en) 2015-08-28 2015-08-28 Evaluation system, evaluation method, and data analysis system

Publications (1)

Publication Number Publication Date
WO2017037768A1 true WO2017037768A1 (en) 2017-03-09

Family

ID=58187088

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2015/074319 WO2017037768A1 (en) 2015-08-28 2015-08-28 Evaluation system, evaluation method, and data analysis system

Country Status (2)

Country Link
JP (1) JP6326510B2 (en)
WO (1) WO2017037768A1 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018061249A1 (en) * 2016-09-30 2018-04-05 株式会社日立製作所 Marketing assistance system
WO2018225227A1 (en) * 2017-06-08 2018-12-13 富士通株式会社 Information processing method, information processing device, and information processing computer program
CN109684468A (en) * 2018-12-13 2019-04-26 四川大学 For the document screening mark platform of evidence-based medicine EBM
CN114080792A (en) * 2019-06-21 2022-02-22 Ntt通信公司 Guidance target evaluation device, guidance target evaluation method, and program

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004029971A (en) * 2002-06-21 2004-01-29 Fujitsu Ltd Data analyzing method
JP2005148806A (en) * 2003-11-11 2005-06-09 Hitachi Ltd Stock operation support system and method
JP2007329415A (en) * 2006-06-09 2007-12-20 Fujitsu Ltd Data processing method, data processing program, recording medium recording same program, and data processor
JP2014081750A (en) * 2012-10-16 2014-05-08 Hitachi Ltd Data integration and analysis system
JP5692091B2 (en) * 2009-12-15 2015-04-01 日本電気株式会社 Information processing apparatus, information processing method, and computer program

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004029971A (en) * 2002-06-21 2004-01-29 Fujitsu Ltd Data analyzing method
JP2005148806A (en) * 2003-11-11 2005-06-09 Hitachi Ltd Stock operation support system and method
JP2007329415A (en) * 2006-06-09 2007-12-20 Fujitsu Ltd Data processing method, data processing program, recording medium recording same program, and data processor
JP5692091B2 (en) * 2009-12-15 2015-04-01 日本電気株式会社 Information processing apparatus, information processing method, and computer program
JP2014081750A (en) * 2012-10-16 2014-05-08 Hitachi Ltd Data integration and analysis system

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018061249A1 (en) * 2016-09-30 2018-04-05 株式会社日立製作所 Marketing assistance system
WO2018225227A1 (en) * 2017-06-08 2018-12-13 富士通株式会社 Information processing method, information processing device, and information processing computer program
JPWO2018225227A1 (en) * 2017-06-08 2020-04-09 富士通株式会社 Information processing method, information processing apparatus, and information processing computer program
CN109684468A (en) * 2018-12-13 2019-04-26 四川大学 For the document screening mark platform of evidence-based medicine EBM
CN109684468B (en) * 2018-12-13 2023-05-09 四川大学 Document screening and labeling system aiming at evidence-based medicine
CN114080792A (en) * 2019-06-21 2022-02-22 Ntt通信公司 Guidance target evaluation device, guidance target evaluation method, and program
US12015546B2 (en) 2019-06-21 2024-06-18 Ntt Communications Corporation Routing destination evaluation apparatus, routing destination evaluating method and program

Also Published As

Publication number Publication date
JP6326510B2 (en) 2018-05-16
JPWO2017037768A1 (en) 2017-08-31

Similar Documents

Publication Publication Date Title
Ripperda et al. Cost effects of modular product family structures: Methods and quantification of impacts to support decision making
Zur Mühlen et al. Business process analytics
Mahendrawathi et al. Analysis of customer fulfilment with process mining: A case study in a telecommunication company
JP6326510B2 (en) Evaluation system, evaluation method, and data analysis system
US9208209B1 (en) Techniques for monitoring transformation techniques using control charts
US20150149257A1 (en) Systems and methods for enterprise profit optimization
AU2020203862A1 (en) Artificial intelligence (ai) based predictions and recommendations for equipment
JP2013080456A (en) System, method, and program for enterprise evaluation
Jovanovic et al. Manufacturing cycle time analysis and scheduling to optimize its duration
Siregar et al. Forecasting of raw material needed for plastic products based in income data using ARIMA method
Stricker et al. Supporting multi-level and robust production planning and execution
Dachyar et al. Designing process improvement of finished good on time release and performance indicator tool in milk industry using business process reengineering method
Uzochukwu et al. Lean production: A frontier for improving performance of oil and gas companies in Nigeria
Araman et al. A case study on implementing Lean Six Sigma: DMAIC methodology in aluminum profiles extrusion process
Lutoshkin et al. The Mathematical Model for Describing the Principles of Enterprise Management “Just in Time, Design to Cost, Risks Management”
Roessler et al. Enhancement of the overall equipment effectiveness measure: a contribution for handling uncertainty in shop floor optimisation and production planning
Zaloha et al. Analysis of the theories for assessment of the quality management product efficiency
Frey et al. Event Entry Time Prediction in Financial Business Processes Using Machine Learning-A Use Case From Loan Applications
Patel et al. Literature search consisting of the areas of Lean Six Sigma usage in Indian SMEs
KR102157216B1 (en) Apparatus and method for providing asset management service
Li et al. Multi-criteria performance management methodology for decision support in industrial project selection problems
Schobel et al. Business process intelligence tools
Lee et al. Asset lifecycle information quality management: a six-sigma approach
US20130061201A1 (en) System and method for determining defect trends
CN117891859B (en) Data processing method and system for industrial Internet of things

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2016574300

Country of ref document: JP

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15902894

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15902894

Country of ref document: EP

Kind code of ref document: A1