CN117196105A - People number prediction method, device, computer equipment and storage medium - Google Patents
People number prediction method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN117196105A CN117196105A CN202311167665.7A CN202311167665A CN117196105A CN 117196105 A CN117196105 A CN 117196105A CN 202311167665 A CN202311167665 A CN 202311167665A CN 117196105 A CN117196105 A CN 117196105A
- Authority
- CN
- China
- Prior art keywords
- machine learning
- preset
- people
- data
- algorithm
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 72
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 382
- 238000010801 machine learning Methods 0.000 claims abstract description 338
- 238000013528 artificial neural network Methods 0.000 claims description 64
- 238000002372 labelling Methods 0.000 claims description 61
- 230000008859 change Effects 0.000 claims description 54
- 238000007637 random forest analysis Methods 0.000 claims description 53
- 238000012549 training Methods 0.000 claims description 42
- 238000004458 analytical method Methods 0.000 claims description 31
- 230000000737 periodic effect Effects 0.000 claims description 28
- 238000004590 computer program Methods 0.000 claims description 24
- 238000012216 screening Methods 0.000 claims description 24
- 238000012545 processing Methods 0.000 claims description 18
- 238000005311 autocorrelation function Methods 0.000 claims description 17
- 230000008569 process Effects 0.000 description 18
- 239000010410 layer Substances 0.000 description 17
- 238000010586 diagram Methods 0.000 description 14
- 238000011156 evaluation Methods 0.000 description 14
- 238000003066 decision tree Methods 0.000 description 13
- 210000002569 neuron Anatomy 0.000 description 10
- 238000007477 logistic regression Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 239000002356 single layer Substances 0.000 description 4
- 230000003442 weekly effect Effects 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000000630 rising effect Effects 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 230000010355 oscillation Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a people number prediction method, a device, computer equipment and a storage medium. The method comprises the following steps: acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time; inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold; and outputting the number of people prediction result at the future time. By adopting the method, the accuracy of people number prediction in the building can be improved.
Description
Technical Field
The application relates to the technical field of big data, in particular to a method and a device for predicting the number of people, computer equipment and a storage medium.
Background
In recent years, with the continuous increase of requirements of building operation management, it is becoming more important to develop reliable operation management strategies for future time periods. In order to make an operation management policy for a future time zone, it is necessary to consider the number of people in the building for the future time zone, and thus it is necessary to predict the number of people in the building for the future time zone.
At present, many researches on methods for predicting the number of people in a building appear, for example, methods such as a mathematical model based on probability statistics, a prediction model based on a Markov chain and the like can be adopted for predicting the number of people in the building.
However, when the number of people in a building is predicted by adopting the traditional number of people prediction method, the problem of lower number of people prediction precision still exists.
Disclosure of Invention
In view of the above, it is desirable to provide a method, an apparatus, a computer device, and a computer-readable storage medium for predicting the number of persons in a building, which can improve the prediction accuracy in predicting the number of persons in the building.
In a first aspect, the present application provides a method of people prediction. The method comprises the following steps:
acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
And outputting the number of people prediction result at the future time.
In one embodiment, the timing characteristics include correlation characteristics and change rule characteristics; the time sequence feature of the actual number of people data in the preset building is obtained in the preset time period, and the time sequence feature comprises the following steps:
acquiring actual number of people data in a preset building in a preset time period;
carrying out standardized processing on the actual people number data to obtain personnel work and rest information corresponding to the actual people number data;
and extracting the correlation characteristics and the change rule characteristics of the actual people number data from the personnel work and rest information corresponding to the actual people number data.
In one embodiment, the correlation features include a first correlation feature and a second correlation feature; the change rule features comprise change trend features and periodic features; extracting the correlation characteristic and the change rule characteristic of the people number data from the people work and rest information corresponding to the actual people number data comprises the following steps:
performing autocorrelation analysis on the personnel work and rest information by adopting an autocorrelation function to obtain the first correlation characteristic;
performing partial autocorrelation analysis on the personnel work and rest information by adopting a partial autocorrelation function to obtain the second correlation characteristic;
And carrying out change trend analysis and periodic analysis on the personnel work and rest information to obtain the change trend characteristics and the periodic characteristics.
In one embodiment, the method further comprises:
acquiring a training sample set and a labeling sample set; the training sample set comprises historical people number data in the preset building at a plurality of preset historical moments, and the labeling sample set comprises labeling people number data at future moments corresponding to the plurality of preset historical moments;
acquiring time sequence characteristics of the historical number data aiming at the historical number data in the preset building at each preset historical time;
inputting the time sequence characteristics of each historical population data into an initial machine learning model for training to obtain predicted population data of future moments corresponding to the preset historical moments; the initial machine learning model comprises at least two initial machine learning algorithms;
and processing the at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
In one embodiment, the processing the at least two initial machine learning algorithms according to the labeling people number data and the predicting people number data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm includes:
updating parameters of the at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain at least two preset machine learning algorithms;
and screening a target machine learning algorithm from the at least two preset machine learning algorithms, and obtaining a preset machine learning model based on the target machine learning algorithm.
In one embodiment, the at least two initial machine learning algorithms include a random forest algorithm and a fully connected neural network algorithm, and the predicted population data includes first predicted population data corresponding to the random forest algorithm and second predicted population data corresponding to the fully connected neural network algorithm; the updating of parameters of the at least two initial machine learning algorithms according to the labeling number of people data and the predicting number of people data to obtain at least two preset machine learning algorithms comprises:
Updating parameters of a random forest algorithm according to the labeling number data and the first predicted number data to obtain a preset random forest algorithm;
and updating parameters of the full-connection neural network algorithm according to the labeling number data and the second predicted number data to obtain a preset full-connection neural network algorithm.
In one embodiment, the screening the target machine learning algorithm from the at least two preset machine learning algorithms, and obtaining the preset machine learning model based on the target machine learning algorithm includes:
calculating a first error between the first predicted population data and the annotated population data;
calculating a second error between the second predicted population data and the annotated population data;
screening out preset machine learning algorithms with errors smaller than a preset error threshold value from the at least two preset machine learning algorithms according to the first errors and the second errors, taking the preset machine learning algorithm with the errors smaller than the preset error threshold value as a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
In a second aspect, the application also provides a people number prediction device. The device comprises:
The time sequence feature acquisition module is used for acquiring time sequence features of actual number data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
the people number prediction module is used for inputting the time sequence characteristics of the actual people number data into a preset machine learning model to perform people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
and the prediction result output module is used for outputting the number of people prediction result at the future moment.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the steps of the method according to any of the first aspects above when the processor executes the computer program.
In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method of any of the first aspects above.
The people number prediction method, the device, the computer equipment and the storage medium acquire the time sequence characteristics of the actual people number data in the preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time; inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold; and outputting the number of people prediction result at the future time. The application can acquire the time sequence characteristics of the actual number of people data in the preset building from the current moment and at least one historical moment adjacent to the current moment. When the time sequence characteristics are acquired, not only the time sequence characteristics of the actual number of people in the preset building at the current moment are considered, but also the time sequence characteristics of the actual number of people in the preset building at least one historical moment adjacent to the current moment are considered, so that the acquired time sequence characteristics are accurate and comprehensive. Secondly, the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms, and the error of the target machine learning algorithm is smaller than a preset error threshold. Obviously, the target machine learning algorithm may be any preset machine learning algorithm with a smaller error of at least two preset machine learning algorithms, and thus, the target machine learning algorithm is not fixed as a single algorithm model. Then, the time sequence characteristics of the actual people number data are input into a target machine learning algorithm obtained through screening to predict, so that the accuracy of people number prediction in the building can be greatly improved.
Drawings
FIG. 1 is a diagram of an application environment for a people prediction method in one embodiment;
FIG. 2 is a flow chart of a method for predicting a number of people in one embodiment;
FIG. 3 is a flowchart illustrating steps for acquiring time series characteristics of actual people number data in a preset building according to one embodiment;
FIG. 4 is a schematic diagram of the decomposition of personnel information in one embodiment;
FIG. 5 is a flow chart of extracting correlation features and change rule features of people data according to an embodiment;
FIG. 6 is a schematic diagram of autocorrelation coefficients of current temporal head count data and historical temporal head count data in one embodiment;
FIG. 7 is a schematic diagram of the partial autocorrelation coefficients of the current time count data and the historical time count data in one embodiment;
FIG. 8 is a flow diagram of a model training process in one embodiment;
FIG. 9 is a schematic diagram of an algorithm flow of a random forest algorithm in one embodiment;
FIG. 10 is a schematic flow diagram of an algorithm of a fully connected neural network algorithm in one embodiment;
FIG. 11 is a flowchart of acquiring a preset machine learning model according to an embodiment;
FIG. 12 is a flow chart of a method of predicting a number of people in an exemplary embodiment;
FIG. 13 is a schematic diagram of various stages of a method of predicting a number of persons in an exemplary embodiment;
FIG. 14 is a schematic diagram of a process of algorithm evaluation in a model training process in an exemplary embodiment;
FIG. 15 is a schematic view of a people prediction device in one embodiment;
FIG. 16 is an internal block diagram of a server in one embodiment;
fig. 17 is an internal structural view of a terminal in one embodiment.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
The people number prediction method provided by the embodiment of the application can be applied to an application environment shown in figure 1. The application environment comprises a computer device 102, wherein the computer device 102 acquires time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time; inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold; and outputting the number of people prediction results at the future time. The computer device 102 may be a terminal device, including but not limited to various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, etc. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The computer device 102 may also be implemented as a stand-alone server or as a cluster of servers.
In one embodiment, as shown in fig. 2, a method for predicting the number of people is provided, and the method is applied to the computer device 102 in fig. 1 for illustration, and includes the following steps:
step 202, acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time.
Here, the preset building includes a commercial building, a residential building, or the like, to which the present application is not limited. The preset time period includes a current time and at least one historical time adjacent to the current time, for example, the current time is 2023, 6, 18, 12, 00am, and then the at least one historical time adjacent to the current time includes 2023, 6, 18, 9, 00am, 2023, 6, 17, 12, 00am, 2023, 16, 9, 00am, 2023, 6, 16, 12, 00am, etc.; of course, the application is not limited in this regard.
The time sequence feature is a time sequence feature, and the time sequence feature of the actual people number data is a corresponding relation between the actual people number data and time, namely a feature representing how the corresponding actual people number data changes along with the change of time. The time sequence characteristics of the actual person number data comprise correlation characteristics and change rule characteristics. The correlation characteristic of the actual people number data refers to the correlation of the corresponding actual people number data at different moments, for example, the correlation of the people number data of the 9 th point today and the people number data of the 9 th point yesterday, and the correlation of the people number data of the 10 th point today and the people number data of the 11 th point today; of course, the application is not limited in this regard. The change rule feature of the actual person number data refers to whether the change of the actual person number data follows a certain rule, such as a periodicity rule, a trend rule, and the like, along with the change of time. The periodic law of the actual people data refers to a law that the actual people data changes periodically, for example, people data of 9 points on monday a week shows periodic characteristics, and people data of 18 points on friday a week also shows periodic characteristics. The trend rule of the actual people data refers to the trend of the actual people data, for example, the people data from 9 to 12 points today shows a gradually rising trend, and the people data from 18 to 21 points yesterday shows a gradually falling trend.
Optionally, the computer device obtains the time sequence characteristics of the actual number of people data in the preset building in a preset time period. First, in a preset time period, the computer device acquires actual number of people data in a preset building. For example, the preset time period includes a current time and at least one historical time adjacent to the current time, if the current time is 2023, 6, 18, 12:00 am), then at least one historical time adjacent to the current time comprises 18 days 12 from 2023, 6, 18: 00am pushes forward one week to 2023, 6, 11, 12: all whole point moments within 00 am. Then, at 18 days 12 from 2023, 6 months: 00am pushes forward one week to 2023, 6, 11, 12: and acquiring actual number data in a preset building at each whole point time by the computer equipment at all whole point time in 00 am.
Next, the computer device extracts a time series feature of the actual person number data from the acquired actual person number data. The obtained actual person number data may be preprocessed by the computer device before extracting the time series characteristics of the actual person number data from the obtained actual person number data. Here, the preprocessing may include performing data cleansing, data conversion, data integration, data specification, and the like on the data, wherein the data cleansing includes performing deletion, merging, and the like on abnormal values, repeated values, and the like in the data; the data conversion includes converting the range, format, etc. of the data; the data integration is to organically integrate the data with different sources, formats and characteristic properties logically or physically; the data protocol is to reduce the data volume on the premise of keeping the original appearance of the data as much as possible. Of course, the application is not limited in this regard. Then, the computer device extracts the time sequence characteristics of the actual person number data from the actual person number data obtained after the preprocessing. For example, correlation features, change rule features, and the like of actual person number data are extracted.
Step 204, inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is less than a preset error threshold.
The machine learning algorithm is an algorithm which automatically inducts logic or rules from data and predicts the data according to the inducted logic or rules, and common machine learning algorithms comprise a random forest algorithm, a logistic regression algorithm, a support vector machine algorithm, a full-connection neural network algorithm and the like; of course, the application is not limited in this regard.
Optionally, the preset machine learning model includes a target machine learning algorithm selected from at least two preset machine learning algorithms, and an error of the target machine learning algorithm is less than a preset error threshold. In other words, the computer device may perform the head count prediction according to at least two preset machine learning algorithms to obtain a head count prediction result, and screen out a preset machine learning algorithm corresponding to the head count prediction result with a smaller error from the head count prediction result, as the target machine learning algorithm.
And then, the computer equipment inputs the time sequence characteristics of the actual people number data into a preset machine learning model to conduct people number prediction, and a people number prediction result of a future moment corresponding to the current moment is obtained. Optionally, inputting the time sequence characteristics of the actual people number data into a target machine learning algorithm for people number prediction, and obtaining a people number prediction result at a future time corresponding to the current time.
If the target machine learning algorithm is a random forest algorithm, inputting the time sequence characteristics of the actual people number data into the random forest algorithm for people number prediction; if the target machine learning algorithm is a logistic regression algorithm, inputting the time sequence characteristics of the actual people number data into the logistic regression algorithm to predict the people number; if the target machine learning algorithm is a fully-connected neural network algorithm, inputting the time sequence characteristics of the actual people number data into the fully-connected neural network algorithm for people number prediction; of course, the application is not limited in this regard.
And 206, outputting the number of people prediction result at the future time.
Where future time refers to the time to be predicted. For example, if the current time is 2023, 6, 18, 12:00am, then the future time includes 2023, 6, 18, 12: all full or partial moments within a week after 00 am. The predicted number of people at the future time may refer to predicted number of people data at the future time.
Optionally, the head count prediction result in step 204 is output as a head count prediction result at a future time. Further, the computer equipment adjusts the operation management strategy of the preset building based on the number of people prediction results at the future time, and optimizes the current building operation management strategy so as to achieve the aim of building energy conservation.
In the embodiment of the application, the computer equipment acquires the time sequence characteristics of the actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time; inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold; and outputting the number of people prediction results at the future time. And acquiring the time sequence characteristics of the actual number of people data in the preset building from the current moment and at least one historical moment adjacent to the current moment. When the time sequence characteristics are acquired, not only the time sequence characteristics of the actual number of people in the preset building at the current moment are considered, but also the time sequence characteristics of the actual number of people in the preset building at least one historical moment adjacent to the current moment are considered, so that the acquired time sequence characteristics are accurate and comprehensive. Secondly, the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms, and the error of the target machine learning algorithm is smaller than a preset error threshold. Obviously, the target machine learning algorithm may be any preset machine learning algorithm with a smaller error of at least two preset machine learning algorithms. Then, when the time sequence characteristics of the actual people number data are input into the target machine learning algorithm obtained through screening for prediction, the accuracy of the prediction of the people number in the building can be improved.
In the above embodiment, the process of obtaining the time sequence characteristics of the actual people number data in the preset building and inputting the time sequence characteristics of the actual people number data into the target machine learning algorithm obtained by screening to predict the people number is involved. In this embodiment, the further description timing characteristics include correlation characteristics and change rule characteristics; in a preset time period, acquiring time sequence characteristics of actual people number data in a preset building, as shown in fig. 3, including:
step 302, acquiring actual number of people data in a preset building in a preset time period.
The actual number of people data in the preset building is the actual number of people data collected in the real environment in the preset building. The preset time period includes a current time and at least one historical time adjacent to the current time, for example, the current time is 2023, 6, 18, 12, 00am, and then the at least one historical time adjacent to the current time includes 2023, 6, 18, 9, 00am, 2023, 6, 17, 12, 00am, 2023, 16, 9, 00am, 2023, 6, 16, 12, 00am, etc.; of course, the application is not limited in this regard.
Optionally, in the preset real environment in the building, the actual people number data is collected at intervals of a fixed time, where the fixed time may be 1 hour or 2 hours, and the method is not limited herein. Collecting and sorting the acquired actual number of people data into a table, and further acquiring the actual number of people data in a preset building in a preset time period.
And 304, carrying out standardization processing on the actual people number data to obtain the people work and rest information corresponding to the actual people number data.
Wherein, the normalization processing of the actual people number data comprises Min-Max normalization (i.e. normalization) processing of the actual people number data.
After the computer equipment acquires the actual number of people data in the preset building in the preset time period, taking the difference of orders of magnitude of the actual number of people data at different moments in the preset building into consideration, the Min-Max standardization (i.e. normalization) processing is required to be carried out on the actual number of people data. The actual people number data is unified into a numerical value which is more than or equal to-1 and less than or equal to 1 in a data conversion mode. And according to the actual people number data at each moment in the preset time period, the personnel work and rest information corresponding to the actual people number data at each moment in the preset time period is obtained.
Step 306, extracting correlation features and change rule features of the actual people number data from the personnel work and rest information corresponding to the actual people number data.
Optionally, according to the personnel information corresponding to the actual personnel number data, analyzing the correlation between the personnel number data at the current moment and the personnel number data at least one historical moment adjacent to the current moment in the personnel information, and obtaining the correlation characteristics of the actual personnel number data. And decomposing the personnel information according to the personnel information corresponding to the actual number of people data to obtain the change rule characteristics of the actual number of people data. Fig. 4 is a schematic diagram of a person work and rest curve, a trend term curve, a period term curve and a random term curve, which are obtained by analyzing actual person number data in a preset time period according to an embodiment. For example, the preset time period may include a time period from 2017, 5, and 2017, 5, and 12, and the preset time period includes a person work and rest curve, a trend term curve, a period term curve, and a random term curve corresponding to actual person number data from 2017, 5, and 2017, 5, and 12, in order from top to bottom in fig. 4. The abscissa of the personnel work and rest curve, the trend term curve, the period term curve and the random term curve all represent time, and the ordinate of the personnel work and rest curve, the trend term curve, the period term curve and the random term curve all represent standardized actual people number data.
Optionally, the correlation of the actual person number data at the current moment and the actual person number data at least one historical moment adjacent to the current moment can be analyzed from the person work and rest curve, and the correlation characteristic of the actual person number data can be obtained. The personnel work and rest curves can be decomposed to obtain change rule characteristics of actual personnel number data, for example, a trend term curve, a period term curve and a random term curve corresponding to the personnel work and rest curves are obtained.
In the embodiment of the application, the actual number of people data in a preset building is obtained in a preset time period;
carrying out standardized processing on the actual people number data to obtain the people work and rest information corresponding to the actual people number data; and extracting the correlation characteristics and the change rule characteristics of the actual people number data from the personnel work and rest information corresponding to the actual people number data. Firstly, the actual people number data for extracting the time sequence characteristics is obtained from the real environment, and obviously, the obtained actual people number data is more fit with the actual scene and has high accuracy. And secondly, when the time sequence characteristics of the actual number of people data in the preset building are acquired, acquiring the time sequence characteristics from two dimensions of the correlation characteristics and the change rule characteristics. The correlation characteristic reflects the correlation between the number of people data at the current moment and at least one historical moment adjacent to the current moment; the change rule features show the change trend or the periodicity rule and the randomness rule of the actual people number data. Thus, the acquired timing characteristics are more comprehensive. Finally, inputting the time sequence characteristics of accurate and comprehensive actual people number data into a preset machine learning model for prediction, so that the precision of people number prediction can be improved.
In the above embodiment, a process of extracting a time series feature of actual person number data is described. In this embodiment, the further description of the correlation features includes a first correlation feature and a second correlation feature; the change rule features include change trend features and periodic features. Then, as shown in fig. 5, extracting the correlation feature and the change rule feature of the actual people number data from the people work and rest information corresponding to the actual people number data includes:
step 502, performing autocorrelation analysis on the personnel work and rest information by adopting an autocorrelation function to obtain a first correlation characteristic.
Wherein the autocorrelation function is used to describe the correlation of the data itself at different times.
Optionally, for example, x (t) represents actual people number data corresponding to time t in a preset time period, and the value range of t is [1, N ]]X (t) corresponding to different t values forms a sequence X t 。
First, X is calculated using the following formula (1) t Is the average value of (a):
μ=E(X t ) (1)
mu in the formula (1) is X t Is a mean value of (c).
Then the following formula (2) is adopted to calculate X t Is the variance of:
σ 2 =D(X t )=E((X t -μ) 2 ) (2)
sigma in formula (2) 2 Namely X t Is a variance of (c).
Recalculating X t Where X is calculated t The auto-covariance of (c) can be divided into two cases, namely unbiased estimation and biased estimation. Wherein an unbiased estimate is an unbiased inference when the sample estimator is used to estimate the overall parameter. If the mathematical expectation of the sample estimator is equal to the true value of the estimated overall parameter, then the sample estimator is referred to as an unbiased estimate of the estimated overall parameter. Biased estimation refers to a biased inference that the sample estimator has systematic errors with the estimated overall parameters, in which case the mathematical expectation of the sample estimator is not equal to the true value of the estimated overall parameters.
The following formula (3) can be used for X t Is used for unbiased estimation:
in the formula (3), k is a variable, C k It can be understood that the covariance of the actual person number data x (t-k) at the time (t-k) and the actual person number data x (t) at the time t, the covariance of x (t-k) and x (t) describing the overall error of x (t-k) and x (t).
The following formula (4) can be used for X t Biased estimation is performed by the auto-covariance of (c):
in the formula (4), k is a variable, C k It can be understood that the covariance of the actual person number data x (t-k) at the time (t-k) and the actual person number data x (t) at the time t, the covariance of x (t-k) and x (t) describing the overall error of x (t-k) and x (t).
Then, the auto-covariance in the unbiased estimation and the auto-covariance in the biased estimation are obtained by calculation. Based on the ratio of the auto-covariance to the variance in unbiased estimation, X is obtained t Autocorrelation coefficient, X, in unbiased estimation t Autocorrelation coefficients at biased estimation. Here, for X t Autocorrelation in unbiased estimationNumber and/or X t The autocorrelation coefficients at the biased estimate are the first correlation features. Fig. 6 is a schematic diagram showing an autocorrelation coefficient of the current time of day people data and the historical time of day people data within a certain 72 hours in one embodiment. In fig. 6, the abscissa indicates time, the ordinate indicates an autocorrelation coefficient, and the closer the value of the autocorrelation coefficient is to 1, the stronger the autocorrelation between the current-time head count data and the historical-time head count data is indicated.
And step 504, performing partial autocorrelation analysis on the personnel work and rest information by adopting a partial autocorrelation function to obtain a second correlation characteristic.
Wherein the partial autocorrelation function is used to describe the direct correlation of data at different times. For example, x (t) represents actual person number data corresponding to time t, and the value range of t is [1, N ]]X (t) corresponding to different t values forms a sequence X t . the actual person number data x (t) corresponding to the time t and the actual person number data x (t-k) corresponding to the time (t-k) have correlation, and the partial autocorrelation coefficient is used to describe the direct influence of x (t-k) on x (t), without considering other cases. For example, the actual person number data x (t-k) corresponding to the time (t-k) affects the actual person number data x (t-k+1) corresponding to the time (t-k+1), and the actual person number data x (t) corresponding to the time (t-k+1) affects the actual person number data x (t) corresponding to the time t, indirectly forming the effect of the actual person number data x (t-k) corresponding to the time (t-k) on the actual person number data x (t) corresponding to the time t. The sequence formed by the partial autocorrelation coefficients is the partial autocorrelation function.
Alternatively, for example, the calculation formula of the actual person number data x (k+1) corresponding to the time (k+1) is shown in the following formula (5):
x(k+1)=φ 1 x(k)+φ 2 x(k-1)+…+φ p x(k-p+1)+ξ k+1 (5)
In the formula (5), phi 1 、φ 2 …φ p Is a linear correlation coefficient, ζ k+1 Is noise phi p That is, as the partial autocorrelation coefficient to be solved, a least square method may be generally used to solve the partial autocorrelation coefficient. Taking p=1 as an example, when p=1, the actual person number data x (k+1) corresponding to the time (k+1) is described asThe calculation formula is shown in the following formula (6):
x(k+1)=φ 1 x(k)+ξ k+1 (6)
the solution formula using the least square method is shown in the following formula (7):
phi in formula (7) 1 The partial autocorrelation coefficient when p=1 is the partial autocorrelation coefficient obtained is the second correlation characteristic. As shown in fig. 7, the partial autocorrelation coefficients of the current time people data and the historical time people data within 72 hours of the people information are selected, in fig. 7, the abscissa represents time, the ordinate represents the partial autocorrelation coefficients, and the closer the value of the partial autocorrelation coefficients is to 1, the stronger the partial autocorrelation between the current time people data and the historical time people data is.
And step 506, carrying out change trend analysis and periodic analysis on the personnel work and rest information to obtain change trend characteristics and periodic characteristics.
The trend feature is a trend of actual people data, for example, people data from 9 to 12 points today shows a gradually rising trend, and people data from 18 to 21 points yesterday shows a gradually falling trend. Periodic characteristics refer to the fact that the actual people data changes periodically, for example, 9 people data on monday weekly shows periodic characteristics, and 18 people data on friday weekly also shows periodic characteristics.
Optionally, decomposing the personnel work and rest information corresponding to the actual number of people data to obtain periodic item features, trend item features and randomness features corresponding to the actual number of people data, and obtaining variation trend features and periodic features of the actual number of people data according to the periodic item features, trend item features and randomness features corresponding to the actual number of people data.
In the embodiment of the application, the first correlation characteristic is obtained by carrying out autocorrelation analysis on the personnel work and rest information by adopting an autocorrelation function; performing partial autocorrelation analysis on the personnel work and rest information by adopting a partial autocorrelation function to obtain a second correlation characteristic; and carrying out change trend analysis and periodic analysis on the personnel work and rest information to obtain change trend characteristics and periodic characteristics corresponding to the actual number of people data. The first correlation characteristic, the second correlation characteristic, the change trend characteristic and the periodic characteristic are extracted from the actual people number data, the actual people number data is quantitatively depicted from different angles, and further more accurate and comprehensive time sequence characteristics are obtained from the actual people number data. Finally, the accurate and comprehensive time sequence features are input into a preset machine learning model for prediction, so that the accuracy of people number prediction can be greatly improved.
In the above embodiment, the process of extracting the first correlation feature, the second correlation feature, the variation trend feature, and the periodicity feature of the actual person number data is described. In this embodiment, the model training process is further described, as shown in fig. 8, including:
step 802, obtaining a training sample set and a labeling sample set; the training sample set comprises historical number data in a preset building at a plurality of preset historical moments, and the labeling sample set comprises labeling number data at future moments corresponding to the preset historical moments.
The training sample set is a set of historical people number data in a preset building at a plurality of historical moments, is a data sample for simulating fitting, and is used for carrying out gradient descent and learning on training errors in the training process, so that the training errors are continuously reduced. The labeling sample set is a set of labeling people number data at future moments corresponding to a plurality of historical moments, and is used for verifying the prediction results of the model trained by the training sample set. The model training problems can be found out in time by verifying the prediction result of the model trained by the training sample set through the labeling sample set.
Step 804, acquiring time sequence characteristics of the historical number data aiming at the historical number data in the preset building at each preset historical moment.
The time sequence features comprise correlation features and change rule features. The correlation feature refers to the correlation of the actual people data corresponding to different moments, for example, the correlation of the people data of 9 points today and the people data of 9 points yesterday, and the correlation of the people data of 10 points today and the people data of 11 points today. The change rule features are that whether the actual person number data changes along with time or not follows a certain rule, such as a periodicity rule, a trend rule and the like. The periodicity rule refers to that actual people data changes periodically, for example, people data at 9 points on monday weekly is periodic, people data at 18 points on friday weekly is periodic. The trend rule refers to a trend of actual people data change, for example, people data from 9 to 12 points today shows a gradually rising trend, and people data from 18 to 21 points yesterday shows a gradually falling trend.
Optionally, aiming at the historical number data in the preset building at each preset historical moment in the training sample set, acquiring the time sequence characteristics of the historical number data, namely, acquiring the correlation characteristics and the change rule characteristics of the historical number data. For example, the historical people number data can be subjected to standardized processing to obtain the people work and rest information corresponding to the historical people number data; and extracting correlation characteristics and change rule characteristics of the historical number data from the personnel work and rest information corresponding to the historical number data.
Further, extracting the correlation characteristic and the change rule characteristic of the historical people data from the people work and rest information corresponding to the historical people data, including: performing autocorrelation analysis on the personnel work and rest information corresponding to the historical personnel number data by adopting an autocorrelation function to obtain a first correlation characteristic corresponding to the historical personnel number data; performing partial autocorrelation analysis on the personnel work and rest information corresponding to the historical personnel number data by adopting a partial autocorrelation function to obtain a second correlation characteristic corresponding to the historical personnel number data; and carrying out change trend analysis and periodic analysis on the personnel work and rest information corresponding to the historical number of people data to obtain the change trend characteristic and the periodic characteristic corresponding to the historical number of people data.
Step 806, inputting the time sequence characteristics of each historical people number data into an initial machine learning model for training to obtain predicted people number data of future moments corresponding to a plurality of preset historical moments; the initial machine learning model includes at least two initial machine learning algorithms.
The initial machine learning model comprises at least two initial machine learning algorithms, wherein the at least two initial machine learning algorithms can comprise a random forest algorithm, a logistic regression algorithm, a support vector machine algorithm, a fully connected neural network algorithm and the like; of course, the application is not limited in this regard.
Optionally, if the at least two initial machine learning algorithms include a random forest algorithm, inputting the time sequence characteristics of the historical population data into the initial machine learning model for training to obtain predicted population data of the random forest algorithm at future times corresponding to a plurality of preset historical times; if the at least two initial machine learning algorithms comprise the full-connection neural network algorithm, the time sequence characteristics of the historical population data are input into the initial machine learning model for training, and then predicted population data of the full-connection neural network algorithm at future moments corresponding to a plurality of preset historical moments are obtained.
Step 808, processing at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
The labeling people number data is derived from a labeling sample set and is used for verifying the people number prediction results of at least two initial machine learning algorithms.
Optionally, error evaluation is performed on the head number prediction results of at least two initial machine learning algorithms by using the head number marking data respectively to obtain an error evaluation result, and an initial machine learning algorithm with an error smaller than a preset error threshold value is obtained according to the error evaluation result, wherein the initial machine learning algorithm with the error smaller than the preset error threshold value is the target machine learning algorithm. Then, a preset machine learning model can be obtained based on the target machine learning algorithm.
In the embodiment of the application, a training sample set and a labeling sample set are obtained; the training sample set comprises historical number data in a preset building at a plurality of preset historical moments, and the labeling sample set comprises labeling number data at future moments corresponding to the preset historical moments; aiming at the historical number data in the preset building at each preset historical moment, acquiring the time sequence characteristics of the historical number data; inputting the time sequence characteristics of each historical number of people data into an initial machine learning model for training to obtain a plurality of predicted number of people data at future moments corresponding to preset historical moments; the initial machine learning model comprises at least two initial machine learning algorithms; and processing at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm. Because the actual people number data for extracting the time sequence features is obtained from the real environment, the actual scene is attached and the accuracy is high, the time sequence features of the actual people number data can be obtained by extracting the time sequence features of the actual people number data, the initial machine learning model comprises at least two initial machine learning algorithms, the initial machine learning algorithm with the error smaller than the preset error threshold value is obtained through error evaluation and is used as a target algorithm for people number prediction, and the preset machine learning model is obtained based on the target machine learning algorithm. Then, when the time sequence characteristics of the actual people number data are input into the target machine learning algorithm obtained through screening for prediction, the error is smaller, and the accuracy of people number prediction in the building can be improved.
In the above embodiment, the process of model training is described. In this embodiment, further describing that at least two initial machine learning algorithms are processed according to the labeling number of people data and the predicting number of people data to obtain a target machine learning algorithm, and a preset machine learning model is obtained based on the target machine learning algorithm, including:
updating parameters of at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain at least two preset machine learning algorithms;
and screening target machine learning algorithms from at least two preset machine learning algorithms, and obtaining a preset machine learning model based on the target machine learning algorithms.
The labeling people number data is derived from a labeling sample set and is used for verifying the people number prediction results of at least two initial machine learning algorithms. The at least two initial machine learning algorithms may include a random forest algorithm, a logistic regression algorithm, a support vector machine algorithm, a fully connected neural network algorithm, etc.; of course, the application is not limited in this regard.
Optionally, if at least two initial machine learning algorithms include a random forest algorithm, updating parameters of the random forest algorithm according to the labeling number data and the predicted number data corresponding to the random forest algorithm to obtain a preset random forest algorithm; if the at least two initial machine learning algorithms comprise the full-connection neural network algorithm, updating parameters of the full-connection neural network algorithm according to the labeling number data and the predicted number data corresponding to the full-connection neural network algorithm to obtain a preset full-connection neural network algorithm. And then, carrying out error evaluation on the head number prediction results of at least two initial machine learning algorithms by using the head number data, and obtaining an initial machine learning algorithm with the error smaller than a preset error threshold according to the error evaluation result, wherein the initial machine learning algorithm with the error smaller than the preset error threshold is the target machine learning algorithm. Finally, a preset machine learning model can be obtained based on the target machine learning algorithm.
In the embodiment of the application, parameters of at least two initial machine learning algorithms are updated according to the labeling number data and the predicting number data to obtain at least two preset machine learning algorithms; and screening target machine learning algorithms from at least two preset machine learning algorithms, and obtaining a preset machine learning model based on the target machine learning algorithms. By updating parameters of at least two initial machine learning algorithms, the prediction precision of the two initial machine learning algorithms is improved, so that the prediction precision of the target machine learning algorithm obtained by screening is higher, and the prediction precision of the number of people is improved.
In the above embodiment, the process of parameter update at model training is described. In this embodiment, it is further described that the at least two initial machine learning algorithms include a random forest algorithm and a fully connected neural network algorithm, and the predicted population data includes first predicted population data corresponding to the random forest algorithm and second predicted population data corresponding to the fully connected neural network algorithm; updating parameters of at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain at least two preset machine learning algorithms, wherein the method comprises the following steps:
Updating parameters of the random forest algorithm according to the labeling number data and the first predicted number data to obtain a preset random forest algorithm;
and updating parameters of the full-connection neural network algorithm according to the labeling number data and the second predicted number data to obtain a preset full-connection neural network algorithm.
The labeling people number data is derived from a labeling sample set and is used for verifying the people number prediction results of at least two initial machine learning algorithms. If the at least two initial machine learning algorithms include a random forest algorithm and a fully connected neural network algorithm, the predicted population data includes first predicted population data corresponding to the random forest algorithm and second predicted population data corresponding to the fully connected neural network algorithm.
Optionally, if the at least two initial machine learning algorithms include a random forest algorithm, updating parameters of the random forest algorithm according to the labeling number data and the first predicted number data to obtain a preset random forest algorithm. For example, as shown in fig. 9, an algorithm flow chart of a random forest algorithm is shown, through which a part of sample data can be randomly selected from an input data set to train a single decision tree, and a large number of decision trees, such as decision tree 1 and decision tree 2 … decision tree n, are obtained after repeated training, and the large number of decision trees form a random forest. For example, the input data set includes X (t-50), X (t-49), X (t-48), X (t-26), X (t-25), X (t-24), X (t-2) and X (t-1), the number of people prediction results of different decision trees can be obtained through a random forest algorithm, and then the number of people prediction results of different decision trees are integrated and solved, namely, the number of people prediction results of the decision trees are averaged to obtain the first predicted number of people data X (t). The prediction effect of the random forest algorithm is related to the number of decision trees, the random forest algorithm is easy to be under-fitted due to the fact that the number of the decision trees is too small, and the accuracy of the random forest algorithm cannot be remarkably improved due to the fact that the number of the decision trees is too large. Therefore, the number of decision trees can be updated in time according to the error result by calculating the error of the first predicted population data and the marked population data, and repeated training is performed, so that a preset random forest algorithm is obtained.
Fig. 10 is a schematic diagram of an algorithm flow of the fully connected neural network algorithm. The full-connection neural network algorithm is a network comprising a multi-layer perceptron structure. Through the input layer of the fully connected neural network, data can be input to the hidden layer through the activation function, then the characteristics of the data can be separated through the hidden layer, and finally the separated characteristics can be output through the output layer. For example, the input data includes: x (t-50), x (t-49), x (t-48), x (t-26), x (t-25), x (t-24), x (t-2) and x (t-1), and calculating the input data through the hidden layer of the fully connected neural network to obtain a second people number prediction result. Therefore, the parameters of the full-connection neural network algorithm can be updated according to the error result by calculating the error of the second predicted population data x (t) and the labeling population data, so as to obtain the preset full-connection neural network algorithm. Parameters of the full-connection neural network algorithm comprise the number of layers of the neural network, the number of single-layer neurons, the type of optimizers, the discarding rate of the neurons and the like.
In the embodiment of the application, at least two initial machine learning algorithms comprise a random forest algorithm and a fully connected neural network algorithm. Optionally, parameters of the random forest algorithm may be updated according to the labeling number of people data and the first predicted number of people data, to obtain a preset random forest algorithm. Parameters of the full-connection neural network algorithm can be updated according to the labeling number data and the second predicted number data, and the preset full-connection neural network algorithm is obtained. Furthermore, the target machine learning algorithm obtained by screening from the preset random forest algorithm and the preset fully-connected neural network algorithm can be any algorithm with smaller error in the preset random forest algorithm and the preset fully-connected neural network algorithm, so that the target machine learning algorithm is not fixed into a single algorithm model. If the target machine learning algorithm is a random forest algorithm, the accuracy of the target machine learning algorithm is higher because the random forest algorithm adopts an integrated algorithm, and the accuracy of the target machine learning algorithm can be further improved. If the target machine learning algorithm is a fully-connected neural network algorithm, the nonlinear mapping capability and the self-adaption performance of the fully-connected neural network algorithm are strong, so that the nonlinear mapping capability and the self-adaption performance of the target machine learning algorithm can be improved. Finally, the time sequence characteristics of the actual people number data are input into a target machine learning algorithm for prediction, so that the accuracy of the prediction of the people number in the building can be improved from the dimensions of accuracy, nonlinear mapping capability, self-adaption performance and the like.
In the above embodiment, the parameter updating is described for the random forest algorithm and the fully connected neural network algorithm, so as to obtain the preset random forest algorithm and the preset fully connected neural network algorithm. In this embodiment, further describing the process of screening out target machine learning algorithms from at least two preset machine learning algorithms, obtaining a preset machine learning model based on the target machine learning algorithms, as shown in fig. 11, the process of obtaining the preset machine learning model includes:
step 1102, a first error between the first predicted population data and the annotated population data is calculated.
The first predicted population data is population prediction data corresponding to a random forest algorithm, and the first error can be an error between the first predicted population data and the labeled population data. The first error may include a root mean square error, a standard deviation, etc. between the first predicted population data and the annotated population data.
Alternatively, if the first error is a root mean square error, the root mean square error between the first predicted population data and the tagged population data may be calculated using equation (8), where equation (8) is as follows:
in formula (8), x i Indicating the number of persons marked, n indicating the number of first predicted persons, Representing first predicted population data.
And (3) obtaining the root mean square error between the first predicted population data and the marked population data through a formula (8).
Step 1104 calculates a second error between the second predicted population data and the tagged population data.
The second predicted population data is population prediction data corresponding to the fully-connected neural network algorithm, and the second error can be an error between the second predicted population data and the labeled population data. The second error may include a root mean square error, a standard deviation, etc. between the second predicted population data and the annotated population data.
Alternatively, if the second error is a root mean square error, equation (8) may be used to calculate the root mean square error between the second predicted and tagged persons data. When the root mean square error between the second predicted population data and the tagged population data is calculated using equation (8), x i Indicating the number of persons marked, n indicating the number of second predicted persons data,representing second predicted population data.
And (3) obtaining the root mean square error between the second predicted population data and the marked population data through the formula (8).
Step 1106, screening out preset machine learning algorithms with errors smaller than a preset error threshold from at least two preset machine learning algorithms according to the first error and the second error, taking the preset machine learning algorithm with errors smaller than the preset error threshold as a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
If the first error is the root mean square error between the first predicted population data and the marked population data and the second error is the root mean square error between the second predicted population data and the marked population data, the preset machine learning algorithm with the error smaller than the preset error threshold value can be selected from at least two preset machine learning algorithms according to the root mean square error between the first predicted population data and the marked population data and the root mean square error between the second predicted population data and the marked population data.
Alternatively, the error smaller than the preset error threshold may be determined as the first error or the second error by comparing the magnitudes of the first error and the second error. In other words, the smallest error of the first error and the second error is determined by comparing the magnitudes of the first error and the second error. Then, a preset machine learning algorithm with an error smaller than a preset error threshold is used as a target machine learning algorithm. Here, the preset error threshold value may be determined based on an empirical value, which is not limited by the present application.
And if the first error is smaller than the preset error threshold, taking a random forest algorithm corresponding to the first error as a target machine learning algorithm. And if the second error is smaller than the preset error threshold, using a fully connected neural network algorithm corresponding to the second error as a target machine learning algorithm. And finally, obtaining a preset machine learning model based on a target machine learning algorithm.
In the embodiment of the application, first, a first error between first predicted number of people data and marked number of people data is calculated; and calculating a second error between the second predicted population data and the marked population data, and screening preset machine learning algorithms with errors smaller than a preset error threshold value from at least two preset machine learning algorithms according to the first error and the second error, wherein the preset machine learning algorithm with errors smaller than the preset error threshold value is used as a target machine learning algorithm. Obviously, the error of the screened target machine learning algorithm is smaller from at least two preset machine learning algorithms based on the error. Finally, a preset machine learning model is obtained based on a target machine learning algorithm with smaller error, and obviously, the error of the obtained preset machine learning model is smaller. Furthermore, the time sequence characteristics of the actual people number data are input into a target machine learning algorithm which is obtained through screening and has small error for prediction, so that the accuracy of people number prediction in the building can be improved.
In one exemplary embodiment, as shown in fig. 12, a method for predicting a number of persons is provided, applied to a computer device, the method comprising:
step 1202, obtaining a training sample set and a labeling sample set; the training sample set comprises historical number data in a preset building at a plurality of preset historical moments, and the labeling sample set comprises labeling number data at future moments corresponding to the preset historical moments.
Step 1204, acquiring time sequence characteristics of the historical number data for the historical number data in the preset building at each preset historical moment.
Step 1206, inputting the time sequence characteristics of each historical head number data into an initial machine learning model for training to obtain predicted head number data of future moments corresponding to a plurality of preset historical moments; the initial machine learning model includes at least two initial machine learning algorithms.
Step 1208, at least two initial machine learning algorithms include a random forest algorithm and a fully connected neural network algorithm, and the predicted population data includes first predicted population data corresponding to the random forest algorithm and second predicted population data corresponding to the fully connected neural network algorithm; and updating parameters of the random forest algorithm according to the labeling number data and the first predicted number data to obtain a preset random forest algorithm.
And step 1210, updating parameters of the full-connection neural network algorithm according to the labeling number data and the second predicted number data to obtain a preset full-connection neural network algorithm. As shown in table 1 below, are part of the parameters of the fully connected neural network algorithm.
TABLE 1
Parameters (parameters) | Code representation | Updated parameters |
Layer number of neural network | ls | 2 |
Number of single-layer neurons | ns | (109,100) |
Optimizer type | optimizer | 'Adam' |
Neuron discarding rate | dropout | 0.1 |
In table 1, the number of layers of the neural network is used to represent the number of layers of the function mapping layer included in the fully connected neural network, and in this embodiment, the number of layers of the obtained neural network of the preset fully connected neural network algorithm may be 2 layers, which is, of course, not limited in the present application.
The number of single-layer neurons is used for representing the number of neurons corresponding to each layer of neural network in the fully-connected neural network, the number of single-layer neurons is equal to the number of input data, in this embodiment, the number of layers of the neural network in the obtained preset fully-connected neural network algorithm is 2, and the number of neurons corresponding to each layer of the neural network is 109 and 100 respectively.
The optimizer type is used to characterize the type of the optimizer used in the fully-connected neural network, and in this embodiment, the optimizer used in the obtained preset fully-connected neural network algorithm may be an Adam optimizer, which is, of course, not limited in this application. The Adam optimizer is used for updating the variable according to the oscillation condition of the historical gradient and the actual historical gradient after filtering oscillation.
The neuron drop rate (dropout) refers to the probability that neurons remain in the network, which, as a regularization technique, can effectively prevent or reduce the overfitting phenomenon. The labeling people number data is derived from a labeling sample set and is used for verifying the people number prediction results of at least two initial machine learning algorithms.
Step 1212, calculates a first error between the first predicted population data and the tagged population data.
Step 1214, calculating a second error between the second predicted population data and the tagged population data.
In step 1216, a preset machine learning algorithm with an error smaller than a preset error threshold is selected from at least two preset machine learning algorithms according to the first error and the second error, and the preset machine learning algorithm with the error smaller than the preset error threshold is used as a target machine learning algorithm, and a preset machine learning model is obtained based on the target machine learning algorithm.
In step 1218, the actual number of people data in the preset building is obtained during the preset time period.
The actual number of people data in the preset building is the number of people data collected in the real environment in the preset building.
Optionally, in the real environment in the preset building, the collection of the number of people data is performed at intervals of a fixed time, where the fixed time may be 1 hour or 2 hours, and the method is not limited herein. As shown in table 2, the time of the collected data is 0 to 23 hours on the 27 th month of 2020, the time step is 1 hour, as shown in table 1, the number of people on the 6 th month of 2020 is 43, the number of people on the 7 th month of 2020 is 119, and … are not described in detail herein. Collecting and sorting the acquired people number data into a table, and further acquiring actual people number data in a preset building in a preset time period.
TABLE 2
2020-04-27 00:00:00 | 0 |
2020-04-27 01:00:00 | 0 |
2020-04-27 02:00:00 | 0 |
2020-04-27 03:00:00 | 0 |
2020-04-27 04:00:00 | 0 |
2020-04-27 05:00:00 | 0 |
2020-04-27 06:00:00 | 43 |
2020-04-27 07:00:00 | 119 |
2020-04-27 08:00:00 | 459 |
2020-04-27 09:00:00 | 518 |
2020-04-27 10:00:00 | 910 |
2020-04-27 11:00:00 | 1451 |
2020-04-27 12:00:00 | 3072 |
2020-04-27 13:00:00 | 1478 |
2020-04-27 14:00:00 | 1181 |
2020-04-27 15:00:00 | 1012 |
2020-04-27 16:00:00 | 991 |
2020-04-27 17:00:00 | 1085 |
2020-04-27 18:00:00 | 1867 |
2020-04-27 19:00:00 | 1613 |
2020-04-27 20:00:00 | 786 |
2020-04-27 21:00:00 | 389 |
2020-04-27 22:00:00 | 71 |
2020-04-27 23:00:00 | 104 |
Step 1220, the actual people number data is normalized to obtain the people work and rest information corresponding to the actual people number data.
Step 1222, performing autocorrelation analysis on the personnel work and rest information by adopting an autocorrelation function to obtain a first correlation characteristic.
Step 1224, performing partial autocorrelation analysis on the personnel information by using the partial autocorrelation function to obtain a second correlation characteristic.
Step 1226, performing a change trend analysis and a periodicity analysis on the work and rest information of the person to obtain a change trend feature and a periodicity feature.
Step 1228, inputting the first correlation feature, the second correlation feature, the change trend feature and the periodic feature of the actual people number data into a preset machine learning model to perform people number prediction, so as to obtain a people number prediction result at a future time corresponding to the current time; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is less than a preset error threshold.
Step 1230, outputting the predicted result of the number of people at the future time.
The steps 1202-1230 described above may be divided into three phases including a feature analysis phase, a model training phase, and a prediction phase. Wherein the feature analysis stage comprises steps 1202-1204; the model training phase includes steps 1206-1216; the prediction phase includes steps 1218-1230. As shown in fig. 13, the characteristic analysis stage includes a process of acquiring time series characteristics of the historic population data for the historic population data in the preset building at each preset history time. The model training stage comprises the processes of inputting time sequence characteristics of historical head number data, updating parameters of at least two initial machine learning algorithms in the initial machine learning model, carrying out algorithm evaluation on at least two preset machine learning algorithms and outputting predicted head number data. Wherein, the algorithm evaluation of the at least two preset machine learning algorithms refers to a process of screening the target machine learning algorithm from the at least two preset machine learning algorithms. The prediction stage comprises the processes of acquiring actual people number data in a preset building, extracting time sequence characteristics, predicting a preset machine learning model and outputting a people number prediction result at a future moment.
As shown in fig. 14, a process of algorithm evaluation in the model training process. Firstly, inputting time sequence characteristics of historical people number data in a preset building at a preset historical moment into a plurality of preset machine learning algorithms for people number prediction, wherein the plurality of preset machine learning algorithms comprise: presetting a machine learning algorithm 1 and a machine learning algorithm 2 … and presetting a machine learning algorithm n; then, obtaining predicted people number data of future time corresponding to the preset historical time by a plurality of preset machine learning algorithms, wherein the predicted people number data comprises: predicted population data 1, predicted population data 2 … predicted population data n; calculating errors of the at least two preset machine learning algorithms on predicted people number data of future time corresponding to the preset historical time and labeling data of future time corresponding to the preset historical time to obtain a first error and a second error … nth error; and finally, screening a preset machine learning algorithm with the error smaller than a preset error threshold value from the first error and the second error … n-th error, taking the preset machine learning algorithm with the error smaller than the preset error threshold value as a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm. In the embodiment of the application, the actual people number data for extracting the time sequence features is acquired from the real environment, is attached to the actual scene and has high accuracy, the time sequence features of the accurate actual people number data can be obtained by extracting the features of the actual people number data, the time sequence features of the accurate actual people number data are input into the preset machine learning model for prediction, the people number prediction result with high accuracy can be obtained, the preset machine learning model comprises at least two preset machine learning algorithms, the preset machine learning algorithm with the error smaller than the preset error threshold value is obtained through error evaluation and is used as a target algorithm for people number prediction, and the error brought by the preset machine learning model is reduced. In summary, the embodiment of the application predicts the number of people through the accurate actual number of people data characteristics and the target algorithm obtained by screening after error evaluation, thereby improving the accuracy of the number of people prediction.
In the people number prediction method, the time sequence characteristics of the actual people number data in the preset building are obtained from the current time and at least one historical time adjacent to the current time. When the time sequence characteristics are acquired, not only the time sequence characteristics of the actual number of people in the preset building at the current moment are considered, but also the time sequence characteristics of the actual number of people in the preset building at least one historical moment adjacent to the current moment are considered, so that the acquired time sequence characteristics are accurate and comprehensive. Secondly, the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms, and the error of the target machine learning algorithm is smaller than a preset error threshold. Obviously, the target machine learning algorithm may be any preset machine learning algorithm with a smaller error of at least two preset machine learning algorithms, and thus, the target machine learning algorithm is not fixed as a single algorithm model. Then, the time sequence characteristics of the actual people number data are input into a target machine learning algorithm obtained through screening to predict, so that the accuracy of people number prediction in the building can be greatly improved. In addition, since the preset machine learning model comprises at least two preset machine learning algorithms, the number of people can be accurately predicted in different scenes according to the advantages of different preset machine learning algorithms, and therefore the number of people prediction method has wide application prospects in the field of constructional engineering.
It should be understood that, although the steps in the flowcharts related to the above embodiments are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiment of the application also provides a people number prediction device for realizing the above related people number prediction method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitation of the embodiment of the one or more people number prediction devices provided below may be referred to the limitation of the people number prediction method hereinabove, and will not be repeated here.
In one embodiment, as shown in FIG. 15, there is provided a people prediction device 1500, comprising: a timing feature acquisition module 1520, a head number prediction module 1540, and a prediction result output module 1560, wherein:
a time sequence feature acquisition module 1520, configured to acquire a time sequence feature of actual people number data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time.
The people number prediction module 1540 is configured to input a time sequence characteristic of actual people number data into a preset machine learning model to perform people number prediction, so as to obtain a people number prediction result at a future time corresponding to the current time; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is less than a preset error threshold.
The prediction result output module 1560 is configured to output a prediction result of the number of people at a future time.
In one embodiment, the timing characteristics include correlation characteristics and change law characteristics; in a preset time period, acquiring a time sequence feature of actual people number data in a preset building, a time sequence feature acquisition module 1520, including:
The system comprises a people number data acquisition unit, a storage unit and a storage unit, wherein the people number data acquisition unit is used for acquiring actual people number data in a preset building in a preset time period;
the standardized unit is used for carrying out standardized processing on the actual people number data to obtain the personnel work and rest information corresponding to the actual people number data;
the characteristic extraction unit is used for extracting the correlation characteristic and the change rule characteristic of the actual people number data from the personnel work and rest information corresponding to the actual people number data.
In one embodiment, the feature extraction unit is further configured to perform autocorrelation analysis on the personnel information by using an autocorrelation function to obtain a first correlation feature; performing partial autocorrelation analysis on the personnel work and rest information by adopting a partial autocorrelation function to obtain a second correlation characteristic; and carrying out change trend analysis and periodic analysis on the personnel work and rest information to obtain change trend characteristics and periodic characteristics.
In one embodiment, a people prediction device 1500 is provided, further comprising:
the model training module is used for acquiring a training sample set and a labeling sample set; the training sample set comprises historical number data in a preset building at a plurality of preset historical moments, and the labeling sample set comprises labeling number data at future moments corresponding to the preset historical moments; and acquiring time sequence characteristics of the historical number data aiming at the historical number data in the preset building at each preset historical moment.
In one embodiment, the model training module is further configured to input a time sequence feature of each historical population data into an initial machine learning model for training, so as to obtain predicted population data at future times corresponding to a plurality of preset historical times; the initial machine learning model comprises at least two initial machine learning algorithms; and processing at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
In one embodiment, a model training module includes:
and the parameter updating unit is used for updating parameters of at least two initial machine learning algorithms according to the labeling number data and the predicted number data to obtain at least two preset machine learning algorithms.
The algorithm evaluation unit is used for screening target machine learning algorithms from at least two preset machine learning algorithms and obtaining a preset machine learning model based on the target machine learning algorithms.
In one embodiment, the parameter updating unit is further configured to update parameters of the random forest algorithm according to the labeling number of people data and the first predicted number of people data to obtain a preset random forest algorithm;
And updating parameters of the full-connection neural network algorithm according to the labeling number data and the second predicted number data to obtain a preset full-connection neural network algorithm.
In one embodiment, the algorithm evaluation unit is further configured to calculate a first error between the first predicted population data and the annotated population data; calculating a second error between the second predicted population data and the tagged population data; screening out preset machine learning algorithms with errors smaller than a preset error threshold value from at least two preset machine learning algorithms according to the first error and the second error, taking the preset machine learning algorithm with errors smaller than the preset error threshold value as a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
The above-described individual modules in the people prediction device may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 16. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is for storing data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program when executed by a processor implements a method of people prediction.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure thereof may be as shown in fig. 17. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program when executed by a processor implements a method of people prediction. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structures shown in fig. 16 and 17 are merely block diagrams of portions of structures associated with the present inventive arrangements and are not limiting of the computer apparatus to which the present inventive arrangements may be implemented, and that a particular computer apparatus may include more or less components than those shown, or may be combined with certain components, or may have different arrangements of components.
In one embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of:
acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
And outputting the number of people prediction results at the future time.
In an embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
and outputting the number of people prediction results at the future time.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, carries out the steps of the method embodiments described above.
In one embodiment, a computer program product is provided comprising a computer program which, when executed by a processor, performs the steps of:
acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
and outputting the number of people prediction results at the future time.
In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method embodiments described above.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.
Claims (10)
1. A method of predicting a number of people, the method comprising:
acquiring time sequence characteristics of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
inputting the time sequence characteristics of the actual people number data into a preset machine learning model for people number prediction, and obtaining a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
And outputting the number of people prediction result at the future time.
2. The method of claim 1, wherein the timing characteristics include correlation characteristics and change law characteristics; the time sequence feature of the actual number of people data in the preset building is obtained in the preset time period, and the time sequence feature comprises the following steps:
acquiring actual number of people data in a preset building in a preset time period;
carrying out standardized processing on the actual people number data to obtain personnel work and rest information corresponding to the actual people number data;
and extracting the correlation characteristics and the change rule characteristics of the actual people number data from the personnel work and rest information corresponding to the actual people number data.
3. The method of claim 2, wherein the correlation features comprise a first correlation feature and a second correlation feature; the change rule features comprise change trend features and periodic features; extracting the correlation characteristic and the change rule characteristic of the people number data from the people work and rest information corresponding to the actual people number data comprises the following steps:
performing autocorrelation analysis on the personnel work and rest information by adopting an autocorrelation function to obtain the first correlation characteristic;
Performing partial autocorrelation analysis on the personnel work and rest information by adopting a partial autocorrelation function to obtain the second correlation characteristic;
and carrying out change trend analysis and periodic analysis on the personnel work and rest information to obtain the change trend characteristics and the periodic characteristics.
4. A method according to any one of claims 1-3, characterized in that the method further comprises:
acquiring a training sample set and a labeling sample set; the training sample set comprises historical people number data in the preset building at a plurality of preset historical moments, and the labeling sample set comprises labeling people number data at future moments corresponding to the plurality of preset historical moments;
acquiring time sequence characteristics of the historical number data aiming at the historical number data in the preset building at each preset historical time;
inputting the time sequence characteristics of each historical population data into an initial machine learning model for training to obtain predicted population data of future moments corresponding to the preset historical moments; the initial machine learning model comprises at least two initial machine learning algorithms;
and processing the at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
5. The method of claim 4, wherein the processing the at least two initial machine learning algorithms based on the tagged number of people data and the predicted number of people data to obtain a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm, comprises:
updating parameters of the at least two initial machine learning algorithms according to the labeling number data and the predicting number data to obtain at least two preset machine learning algorithms;
and screening a target machine learning algorithm from the at least two preset machine learning algorithms, and obtaining a preset machine learning model based on the target machine learning algorithm.
6. The method of claim 5, wherein the at least two initial machine learning algorithms comprise a random forest algorithm and a fully connected neural network algorithm, the predicted population data comprising first predicted population data corresponding to the random forest algorithm and second predicted population data corresponding to the fully connected neural network algorithm; the updating of parameters of the at least two initial machine learning algorithms according to the labeling number of people data and the predicting number of people data to obtain at least two preset machine learning algorithms comprises:
Updating parameters of a random forest algorithm according to the labeling number data and the first predicted number data to obtain a preset random forest algorithm;
and updating parameters of the full-connection neural network algorithm according to the labeling number data and the second predicted number data to obtain a preset full-connection neural network algorithm.
7. The method of claim 5, wherein the screening the target machine learning algorithm from the at least two preset machine learning algorithms, and deriving a preset machine learning model based on the target machine learning algorithm, comprises:
calculating a first error between the first predicted population data and the annotated population data;
calculating a second error between the second predicted population data and the annotated population data;
screening out preset machine learning algorithms with errors smaller than a preset error threshold value from the at least two preset machine learning algorithms according to the first errors and the second errors, taking the preset machine learning algorithm with the errors smaller than the preset error threshold value as a target machine learning algorithm, and obtaining a preset machine learning model based on the target machine learning algorithm.
8. A people number prediction device, the device comprising:
the time sequence feature acquisition module is used for acquiring time sequence features of actual number of people data in a preset building in a preset time period; the preset time period comprises a current time and at least one historical time adjacent to the current time;
the people number prediction module is used for inputting the time sequence characteristics of the actual people number data into a preset machine learning model to perform people number prediction, so as to obtain a people number prediction result at a future moment corresponding to the current moment; the preset machine learning model comprises a target machine learning algorithm selected from at least two preset machine learning algorithms; the error of the target machine learning algorithm is smaller than a preset error threshold;
and the prediction result output module is used for outputting the number of people prediction result at the future moment.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311167665.7A CN117196105A (en) | 2023-09-11 | 2023-09-11 | People number prediction method, device, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311167665.7A CN117196105A (en) | 2023-09-11 | 2023-09-11 | People number prediction method, device, computer equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117196105A true CN117196105A (en) | 2023-12-08 |
Family
ID=88997453
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311167665.7A Pending CN117196105A (en) | 2023-09-11 | 2023-09-11 | People number prediction method, device, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117196105A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117908003A (en) * | 2024-03-19 | 2024-04-19 | 清澜技术(深圳)有限公司 | Space people counting method and system |
-
2023
- 2023-09-11 CN CN202311167665.7A patent/CN117196105A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117908003A (en) * | 2024-03-19 | 2024-04-19 | 清澜技术(深圳)有限公司 | Space people counting method and system |
CN117908003B (en) * | 2024-03-19 | 2024-06-07 | 清澜技术(深圳)有限公司 | Space people counting method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109816221B (en) | Project risk decision method, apparatus, computer device and storage medium | |
Wu et al. | Evolving RBF neural networks for rainfall prediction using hybrid particle swarm optimization and genetic algorithm | |
CN109034861B (en) | User loss prediction method and device based on mobile terminal log behavior data | |
CN111506723B (en) | Question-answer response method, device, equipment and storage medium | |
Wang et al. | The trend-fuzzy-granulation-based adaptive fuzzy cognitive map for long-term time series forecasting | |
CN112381673B (en) | Park electricity utilization information analysis method and device based on digital twin | |
CN113807568B (en) | Power load prediction method and device and terminal equipment | |
CN117196105A (en) | People number prediction method, device, computer equipment and storage medium | |
CN114065653A (en) | Construction method of power load prediction model and power load prediction method | |
CN112232604A (en) | Prediction method for extracting network traffic based on Prophet model | |
CN116029425A (en) | Method, device and computer equipment for predicting fan output climbing event | |
Qi et al. | An efficient GAN-based predictive framework for multivariate time series anomaly prediction in cloud data centers | |
Syeed et al. | Flood prediction using machine learning models | |
Butt et al. | Hybrid of deep learning and exponential smoothing for enhancing crime forecasting accuracy | |
Liu et al. | Multi-stage residual life prediction of aero-engine based on real-time clustering and combined prediction model | |
CN114254828B (en) | Power load prediction method based on mixed convolution feature extractor and GRU | |
CN116383441A (en) | Community detection method, device, computer equipment and storage medium | |
CN115392567A (en) | Power load prediction method, electronic equipment, device and readable storage medium | |
CN114495137A (en) | Bill abnormity detection model generation method and bill abnormity detection method | |
CN116757338B (en) | Crop yield prediction method, device, electronic equipment and storage medium | |
Astutik et al. | Comparing Different Supervised Machine Learning Accuracy for Weather Analysis on CubeSat Simulation | |
CN114239945B (en) | Short-term power load prediction method, device, equipment and storage medium | |
Mung et al. | Time Series Weather Data Forecasting Using Deep Learning | |
Kraamwinkel | Time Series Forecasting on COVID-19 Data and Its Relevance to International Health Security | |
Sun | LSTM-based Trend Prediction of Public Opinion on Urban Fires |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |