WO2020253381A1 - 数据监控方法、装置、计算机设备和存储介质 - Google Patents
数据监控方法、装置、计算机设备和存储介质 Download PDFInfo
- Publication number
- WO2020253381A1 WO2020253381A1 PCT/CN2020/086914 CN2020086914W WO2020253381A1 WO 2020253381 A1 WO2020253381 A1 WO 2020253381A1 CN 2020086914 W CN2020086914 W CN 2020086914W WO 2020253381 A1 WO2020253381 A1 WO 2020253381A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- monitoring
- data
- initial
- target
- dimension
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Definitions
- This application relates to the field of machine learning technology in artificial intelligence, and in particular to a data monitoring method, device, computer equipment and storage medium.
- a data monitoring method includes:
- Receive data monitoring instructions carry the initial monitoring account ID, initial monitoring dimensions and initial monitoring period;
- a data monitoring device includes:
- the instruction receiving module is used to receive data monitoring instructions.
- the data monitoring instructions carry the initial monitoring account identification, initial monitoring dimension and initial monitoring period;
- the monitoring data obtaining module is used to obtain the initial monitoring data corresponding to the initial monitoring account identifier in the initial monitoring period according to the initial monitoring dimension;
- the probability obtaining module is used to determine the unearned account identification based on the initial monitoring data, and input the initial monitoring data corresponding to the unearned account identification into the trained scoring card model to obtain the scoring result corresponding to the unearned account identification.
- the scoring result is used for Indicates the possibility of overdue corresponding to the unwarned account ID;
- the model calculation module is used to input the initial monitoring data and the scoring result corresponding to the unwarned account identification into the trained naive Bayes model to obtain the output result;
- the monitoring module is used to determine the target monitoring period and the target monitoring dimension according to the output result, and obtain the target monitoring data corresponding to the unwarned account identifier during the target monitoring period according to the target monitoring dimension.
- a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:
- Receive data monitoring instructions carry the initial monitoring account ID, initial monitoring dimensions and initial monitoring period;
- the initial monitoring data corresponding to the initial account identification is acquired during the initial monitoring period
- the target monitoring period and target monitoring dimension are determined according to the output result, and the target monitoring data corresponding to the unwarned account identification is obtained according to the target monitoring period and target monitoring dimension.
- a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:
- Receive data monitoring instructions carry the initial monitoring account ID, initial monitoring dimensions and initial monitoring period;
- the initial monitoring data corresponding to the initial account identification is acquired during the initial monitoring period
- the target monitoring period and target monitoring dimension are determined according to the output result, and the target monitoring data corresponding to the unwarned account identification is obtained according to the target monitoring period and target monitoring dimension.
- the above-mentioned data monitoring method, device, computer equipment and storage medium adjust the initial monitoring period and initial monitoring dimension by using the scorecard model and the naive Bayes model according to the initial monitoring data, thereby obtaining the target monitoring period and target monitoring dimension, and using the target
- the monitoring cycle and target monitoring dimensions are used to obtain target monitoring data corresponding to the target account identifier, which enables different users to monitor using different monitoring cycles and monitoring dimensions, which improves the efficiency of monitoring and saves monitoring resources.
- Figure 1 is an application scenario diagram of a data monitoring method in an embodiment
- Figure 2 is a schematic flow chart of a data monitoring method in an embodiment
- FIG. 3 is a schematic diagram of a process of training a score card model in an embodiment
- FIG. 4 is a schematic diagram of a process of training a naive Bayes model in an embodiment
- Fig. 5 is a schematic diagram of a process for obtaining initial monitoring data in an embodiment
- FIG. 6 is a schematic diagram of a process for obtaining the overdue probability corresponding to an unwarned account identifier in an embodiment
- Figure 7 is a schematic diagram of a visual report in a specific embodiment
- Figure 8 is a structural block diagram of a data monitoring device in an embodiment
- Fig. 9 is an internal structure diagram of a computer device in an embodiment.
- the data monitoring method provided in this application can be applied to the application environment shown in FIG. 1.
- the terminal 102 communicates with the server 104 through the network.
- the server 104 receives the data monitoring instruction sent by the terminal 102.
- the data monitoring instruction carries the initial monitoring account identifier, the initial monitoring dimension, and the initial monitoring period; the server 104 monitors the initial acquisition corresponding to the initial account identifier in the initial monitoring period according to the initial monitoring dimension.
- Monitoring data determine the unwarned account ID according to the initial monitoring data, and input the initial monitoring data corresponding to the unwarned account ID into the trained scoring card model to obtain the scoring result corresponding to the unwarned account ID, and the scoring result is used to indicate the The probability of overdue corresponding to the unwarned account ID; the server 104 inputs the initial monitoring data and scoring results corresponding to the unwarned account ID into the trained naive Bayes model to obtain the output result; the server 104 determines the target according to the output result The monitoring period and the target monitoring dimension, according to the target monitoring dimension, obtain the target monitoring data corresponding to the unwarned account identifier in the target monitoring period.
- the terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices.
- the server 104 may be implemented as an independent server or a server cluster composed of multiple servers.
- a data monitoring method is provided. Taking the method applied to the server in FIG. 1 as an example for description, the method includes the following steps:
- S202 Receive a data monitoring instruction, where the data monitoring instruction carries an initial monitoring account identifier, an initial monitoring dimension, and an initial monitoring period.
- the initial monitoring account identifier is used to uniquely identify the user who needs to be monitored, for example, the loan user to be monitored after the loan is issued by the loan platform.
- the initial monitoring dimension refers to the data fields related to the user corresponding to the initial monitoring account identification to be collected, such as the user's resident address field, the user's mobile phone bill field, and the user's external loan application field.
- the initial monitoring period refers to the preliminarily determined length of time for collecting relevant data of the user, such as one month or two months.
- the server receives a data monitoring instruction sent by the terminal, and the data monitoring instruction carries an initial monitoring account identifier, an initial monitoring dimension, and an initial monitoring period.
- the server parses the data monitoring instruction, and obtains the initial monitoring account identifier, the initial monitoring dimension, and the initial monitoring period carried by the data monitoring instruction.
- S204 Acquire initial monitoring data corresponding to the initial monitoring account identifier in the initial monitoring period according to the initial monitoring dimension.
- the initial monitoring data refers to collecting data corresponding to the initial monitoring account identifier according to the initial monitoring dimension during the initial monitoring period.
- the server monitors the initial monitoring account identifier in the initial monitoring period according to the initial monitoring dimension to obtain the initial monitoring data.
- the server can obtain data corresponding to the initial monitoring dimension of the initial monitoring account identifier from various channels on the Internet.
- the data corresponding to the initial monitoring account identification can also be obtained from the internal database.
- S206 Determine the unearned account identifier according to the initial monitoring data, and input the initial monitoring data corresponding to the unearned account identifier into the trained scoring card model to obtain a scoring result corresponding to the unearned account identifier.
- the scoring result is used to indicate the unearned account identifier.
- the unwarned account identifier is used to identify users who do not meet the preset warning rules, and the unwarned account identifier is included in the initial account identifier.
- the preset warning rules refer to some artificial warning rules. For example, in the initial monitoring data, the user’s mobile phone is in arrears, then the user’s account is identified as the warning account; the user’s resident address is unoccupied for a long time, then The user's account identifier is an early warning account identifier. When the user needs to perform risk processing for the early warning account identification, that is, an early warning prompt is sent to the management terminal and the early warning account identification is displayed, without monitoring the early warning account identification.
- the score card model is a model trained using a machine learning algorithm based on historical monitoring data and scoring results.
- the machine learning algorithm can be a logistic regression algorithm, a deep neural network algorithm, and so on.
- the scorecard model is a model used to predict the user's overdue probability.
- the server determines the early-warning account identifier and the un-alarmed account identifier from the initial monitoring data according to the pre-set early-warning rules, and sends the early-warning account identifier and the warning prompt to the management terminal for display. Then the server inputs the initial monitoring data corresponding to the unwarned account ID into the trained scorecard model, and obtains the scoring result output by the scorecard model.
- the scoring result represents the overdue probability corresponding to the unwarned account ID.
- the naive Bayes model is calculated in advance by using the naive Bayes algorithm based on the historical user loan monitoring data and the score results corresponding to whether the repayment is overdue.
- the output result may be the probability value of the preset monitoring dimension and the preset monitoring period corresponding to each unwarned account identifier.
- the server inputs the initial monitoring data and scoring results corresponding to the unwarned account identifiers into the trained naive Bayes model for calculation, and obtains the probability of the preset monitoring dimensions and the preset monitoring period corresponding to each unwarned account identifier value.
- S210 Determine the target monitoring period and target monitoring dimension according to the output result, and obtain target monitoring data corresponding to the unwarned account identifier during the target monitoring period according to the target monitoring dimension.
- the target monitoring period refers to the period of monitoring the unwarned account identification after the initial monitoring period
- the target monitoring dimension refers to the data fields related to the unwarned account identification that need to be obtained in the target monitoring period.
- Target monitoring data refers to the data corresponding to the target monitoring dimension acquired during the target monitoring period.
- the server uses the preset monitoring dimension and the preset monitoring period that are greater than the preset threshold in the output result as the target monitoring period and the target monitoring dimension corresponding to each unwarned account identification, and the server uses the target monitoring dimension in the target monitoring period according to the target monitoring dimension. Obtain target monitoring data corresponding to each unwarned account identifier within.
- the initial monitoring period and the initial monitoring dimension are adjusted by using the scorecard model and the naive Bayes model according to the initial monitoring data, so as to obtain the target monitoring period and target monitoring dimension, using the target monitoring period and target monitoring dimension
- the target monitoring data corresponding to the target account identification it is realized that different users use different monitoring cycles and monitoring dimensions to monitor, which improves the efficiency of monitoring and saves monitoring resources.
- the step of generating the trained scorecard model includes:
- S302 Obtain historical monitoring data and corresponding overdue results, generate a monitoring data vector according to the historical monitoring data, and generate a corresponding scoring result vector according to the overdue result.
- the scoring result refers to the result of whether the user repays the loan within the specified period after the loan, including overdue and not overdue.
- the server obtains the monitoring data of historical loan users and the corresponding overdue results, and preprocesses the historical monitoring data to obtain the monitoring data vector, that is, to digitize the data corresponding to each dimension in the monitoring data, and then convert the digitized data Is a vector to obtain a vector of monitoring data.
- the overdue result is converted into a scoring result vector.
- the scoring result vector corresponding to the overdue can be (0,1).
- the score result vector corresponding to no warning can be (1,0).
- the monitoring data vector is used as an input, and the scoring result vector corresponding to the monitoring data vector is used as a label for training using a logistic regression algorithm.
- a trained scoring card model is obtained.
- the preset condition means that the number of training times reaches the maximum number of iterations or the value of the used loss function reaches a preset threshold.
- Logistic regression algorithm refers to logistic regression algorithm, which is a generalized linear regression analysis model.
- the activation function uses the sigmoid function
- the loss function uses a maximum likelihood function.
- the server will take the monitoring data vector as input, and use the scoring result corresponding to the monitoring data vector as the label for training using a logistic regression algorithm.
- the result is Trained scorecard model.
- a logistic regression algorithm is used to train the scorecard model through historical monitoring data and corresponding overdue results, which is convenient for subsequent use and improves monitoring efficiency.
- the step of generating the trained naive Bayes model includes:
- S402 Obtain historical monitoring dimensions, corresponding monitoring periods, and scoring results, and calculate non-warning probability parameters and overdue probability parameters based on historical monitoring dimensions, corresponding monitoring periods, and scoring results.
- the unwarned probability parameter refers to the probability parameter of all loan users who have not overdue loans among the historical users
- the overdue probability parameter refers to the probability parameter of all the loan users who have overdue loans among the historical users.
- the server obtains the monitoring dimension, historical monitoring period, and historical scoring results of historical users. According to the historical monitoring dimension, historical monitoring period and historical scoring results, the unwarned probability parameter and the overdue probability parameter are calculated. That is, the overdue probability parameter is obtained according to the number of overdue users and the total number of historical users. According to the total number of users who have not expired and the total number of historical users, the probability parameter of not expired is obtained.
- S404 Calculate the conditional probability parameter of the historical monitoring dimension according to the historical monitoring dimension and the scoring result, calculate the conditional probability parameter of the monitoring period according to the monitoring period and the scoring result, and obtain a trained naive Bayes model.
- the naive Bayes model refers to the model trained using the naive Bayes algorithm, and its naive Bayes formula is The calculation formula to convert this formula to overdue results is
- the server calculates the conditional probability parameter of each dimension in the user's historical monitoring dimension in the overdue result, that is, the scoring result of each dimension in the historical monitoring dimension is the conditional probability parameter of the user overdue, and the scoring result of each dimension in the historical monitoring dimension is the user Conditional probability parameters that are not overdue.
- the monitored field is a mobile phone expense field
- calculate the conditional probability of mobile phone arrears and the conditional probability parameters of mobile phone arrears when the user loan is overdue calculate the conditional probability of mobile phone arrears if the user’s loan is not overdue and The conditional probability parameter of the mobile phone is not in arrears, and the conditional probability parameter of the mobile phone fee field is obtained.
- the conditional probability of the monitoring period is calculated, that is, the conditional probability parameters of different monitoring periods in the case of overdue, and the trained naive Bayes model is obtained according to the calculated parameters.
- the non-warning probability parameter and the overdue probability parameter are calculated according to the historical monitoring dimension, the corresponding monitoring period and the scoring result.
- Calculate the conditional probability parameter of the historical monitoring dimension according to the historical monitoring dimension and the scoring result calculate the conditional probability parameter of the monitoring period according to the monitoring period and the scoring result, and obtain the trained naive Bayes model.
- the naive Bayes model is obtained through pre-training, which can be used directly during monitoring, which is convenient and quick.
- step S204 namely obtaining the initial monitoring data corresponding to the initial monitoring account identifier in the initial monitoring period according to the initial monitoring dimension, includes the steps:
- S502 Obtain a monitoring data message from each preset data source in an initial monitoring period according to the initial monitoring dimension.
- the preset data source refers to a server that can obtain various data corresponding to the initial account identifier. For example, obtain the user's mobile phone expenses from the telecommunication server database, and obtain whether the user is at the resident address from the public security server database. Obtain the loan application status of the user from the server database of other financial institutions and so on.
- the server obtains the monitoring data message corresponding to each initial monitoring account identifier from each preset data source in the initial monitoring period according to the initial monitoring dimension.
- the authorization of each data source needs to be obtained in advance.
- the server parses the initial monitoring data message corresponding to each preset data source to obtain the analysis result. That is, the data corresponding to the corresponding monitoring data message in each preset data source is obtained.
- Each data source has its own message format, and different field data obtained by parsing messages from different data sources.
- the target field refers to a unified field set in advance.
- the target field is used to unify the message fields of each data source so as to obtain data in the same format.
- the server converts the analysis result into data corresponding to the set field according to the preset field, and the initial monitoring data is obtained.
- the initial monitoring data is obtained by converting the analysis results of the data messages of each preset data source into unified field data, so that the fields of the obtained initial monitoring data are unified, which is convenient for subsequent management.
- the target monitoring period and target monitoring dimension are determined according to the output result, and the target monitoring data corresponding to the unwarned account identification is obtained in the target monitoring period according to the target monitoring dimension. After that, it also includes steps:
- S602 Obtain a target monitoring data vector according to the target monitoring data.
- the server preprocesses the target monitoring data, that is, converts it into numerical data, and then obtains the target monitoring data vector according to the numerical target monitoring data.
- S604 Input the target monitoring data vector into the trained scoring card model to obtain the scoring result of the unwarned account identifier in the target monitoring period, and send an alarm prompt to the monitoring terminal according to the scoring result.
- the target monitoring data vector is input into the trained scoring card model, and the scoring result of the unwarned account identification in the target monitoring period is obtained.
- the scoring result is the unwarned account identification mark corresponding to the overdue result as the risk account, and the scoring result corresponding to the risk user is sent to the monitoring terminal and an early warning is given.
- the target monitoring data vector is obtained according to the target monitoring data, and the target monitoring data vector is input into the trained scoring card model to obtain the scoring result of the unwarned account identification in the target monitoring period.
- the terminal sends an alarm reminder, and uses the scorecard model to get the score based on the monitored data, which can promptly remind the monitoring terminal to perform risk control on the overdue account identification in the score result, and avoid overdue repayment of the loan account.
- step S210 after determining the target monitoring period and target monitoring dimension according to the output result, and obtaining target monitoring data corresponding to the unwarned account identifier in the target monitoring period according to the target monitoring dimension, the method further includes the following steps:
- the related monitoring data can be visually displayed for users to view.
- the data monitoring method is applied in a post-loan monitoring and early warning platform.
- the monitoring personnel upload the post-loan customer table to the post-loan monitoring and early warning platform, and select the customer identification and the corresponding initial monitoring field and initial monitoring period in the post-loan customer table to be monitored on the post-loan monitoring and early warning platform.
- the early warning platform obtains the external field data corresponding to the initial monitoring field through Qianhai credit investigation or PBOC credit investigation, etc., converts the external field into internal field through field management, and then performs the initial early warning processing on the data corresponding to the field through rule management.
- Unwarned customer identification input the initial monitoring data corresponding to the unwarned customer identification into the scorecard model to obtain the loan overdue probability corresponding to the unwarned customer identification, and get the overdue result according to the loan overdue probability, that is, the loan overdue probability is greater than the preset threshold
- the customer identification without warning is marked as overdue.
- the post-loan monitoring and early warning platform can perform statistical calculations based on the target monitoring data, obtain a display report, and display the display report on the monitoring terminal.
- the display report can also be sent to the terminal corresponding to the risk control personnel for display, which is convenient for the risk control personnel to deal with risks.
- a data monitoring device 800 including: an instruction receiving module 802, a monitoring data obtaining module 804, a probability obtaining module 806, a model calculation module 808, and a monitoring module 810, wherein:
- the instruction receiving module 802 is used to receive a data monitoring instruction, and the data monitoring instruction carries an initial monitoring account identifier, an initial monitoring dimension, and an initial monitoring period;
- the monitoring data obtaining module 804 is configured to obtain the initial monitoring data corresponding to the initial monitoring account identifier in the initial monitoring period according to the initial monitoring dimension;
- the probability obtaining module 806 is used to determine the unwarned account ID according to the initial monitoring data, and input the initial monitoring data corresponding to the unwarned account ID into the trained scoring card model to obtain the scoring result corresponding to the unwarned account ID.
- the scoring result is used Yu means the possibility of overdue corresponding to the unwarned account identifier;
- the model calculation module 808 is used to input the initial monitoring data and the scoring result corresponding to the unwarned account identifier into the trained naive Bayes model to obtain the output result;
- the monitoring module 810 is configured to determine the target monitoring period and the target monitoring dimension according to the output result, and obtain the target monitoring data corresponding to the unwarned account identifier during the target monitoring period according to the target monitoring dimension.
- the data monitoring device 800 further includes:
- the vector generation module is used to obtain historical monitoring data and corresponding overdue results, generate a monitoring data vector based on the historical monitoring data, and generate a corresponding score result vector based on the overdue result;
- the model training module is used to take the monitoring data vector as an input, and use the scoring result vector corresponding to the monitoring data vector as a label for training using a logistic regression algorithm. When a preset condition is reached, a trained scoring card model is obtained.
- the data monitoring device 800 further includes:
- the parameter calculation module is used to obtain historical monitoring dimensions, corresponding monitoring periods and scoring results, and calculate the unwarned probability parameters and overdue probability parameters according to the historical monitoring dimensions, corresponding monitoring periods and scoring results;
- the conditional parameter calculation module is used to calculate the conditional probability parameter of the historical monitoring dimension according to the historical monitoring dimension and the scoring result, and calculate the conditional probability parameter of the monitoring period according to the monitoring period and the scoring result to obtain the trained naive Bayes model.
- the monitoring data obtaining module 804 includes:
- the data message acquisition module is used to acquire the initial monitoring data message from each preset data source in the initial monitoring period according to the initial monitoring dimension;
- the parsing module is used to parse the initial monitoring data message corresponding to each preset data source to obtain the parsing result;
- the data conversion module is used to convert the analysis result into the data corresponding to the target field to obtain the initial monitoring data.
- the data monitoring device 800 further includes:
- the target vector obtaining module is used to obtain the target monitoring data vector according to the target monitoring data
- the monitoring module is used to input the target monitoring data vector into the trained scoring card model to obtain the scoring result of the unwarned account identification in the target monitoring period, and send an alarm prompt to the monitoring terminal according to the scoring result.
- the data monitoring device 800 further includes:
- the visualization generation module is used to perform statistical calculations based on the target monitoring data to obtain the calculation results, generate visualization reports based on the calculation results, and send the visualization reports to the monitoring terminal for display.
- Each module in the above-mentioned data monitoring device can be implemented in whole or in part by software, hardware, and a combination thereof.
- the foregoing modules may be embedded in the form of hardware or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the foregoing modules.
- a computer device is provided.
- the computer device may be a server, and its internal structure diagram may be as shown in FIG. 9.
- the computer equipment includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide calculation 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, a computer program, and a database.
- the internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
- the computer equipment database is used to store monitoring related data.
- the network interface of the computer device is used to communicate with an external terminal through a network connection.
- the computer program is executed by the processor to realize a data monitoring method.
- the method includes the following steps: receiving a data monitoring instruction, the data monitoring instruction carries an initial monitoring account identifier, an initial monitoring dimension, and an initial monitoring period;
- the initial monitoring dimension acquires the initial monitoring data corresponding to the initial monitoring account identifier during the initial monitoring period; determines the unwarned account identifier according to the initial monitoring data, and inputs the initial monitoring data corresponding to the unearned account identifier to
- the scoring result corresponding to the un-alarmed account identifier is obtained, and the scoring result is used to indicate the overdue probability corresponding to the un-alarmed account identifier;
- the initial The monitoring data and the scoring result are input into the trained naive Bayes model to obtain the output result;
- the target monitoring period and target monitoring dimension are determined according to the output result, and the target monitoring dimension is within the target monitoring period according to the target monitoring dimension
- FIG. 9 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied.
- the specific computer equipment may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.
- a computer device including a memory and a processor, the memory stores a computer program, and the processor implements the steps described in the electronic invoice generation method in any of the foregoing embodiments when the processor executes the computer program.
- a computer-readable storage medium is provided.
- the storage medium is a volatile storage medium or a non-volatile storage medium, and a computer program is stored thereon.
- the computer program is executed by a processor, any of the foregoing is implemented.
- the steps of the electronic invoice generation method in the embodiment include: receiving a data monitoring instruction, the data monitoring instruction carrying an initial monitoring account identifier, an initial monitoring dimension, and an initial monitoring period; according to the initial monitoring dimension in the initial monitoring Acquire the initial monitoring data corresponding to the initial monitoring account identifier during the monitoring period; determine the unearned account identifier according to the initial monitoring data, and input the initial monitoring data corresponding to the unearned account identifier into the trained scoring card model, Obtain the scoring result corresponding to the un-alarmed account identifier, and the scoring result is used to indicate the overdue probability corresponding to the un-alarmed account identifier; input the initial monitoring data corresponding to the un-alarmed account identifier and the scoring result
- the output result is obtained; the target monitoring period and target monitoring dimension are determined according to the output result, and the corresponding unwarned account identification is obtained in the target monitoring period according to the target monitoring dimension Target monitoring data.
- Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
- Volatile memory may include random access memory (RAM) or external cache memory.
- RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Business, Economics & Management (AREA)
- Technology Law (AREA)
- Evolutionary Biology (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Algebra (AREA)
- Development Economics (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
本申请涉及人工智能中的机器学习,提供一种数据监控方法、装置、计算机设备和存储介质。该方法包括:接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;根据初始监控维度在初始监控周期内获取初始账号标识对应的初始监控数据;根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果;将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据。采用本方法能够节省监控资源。
Description
本申请要求于2019年6月17日提交中国专利局、申请号为201910521514.4,发明名称为“数据监控方法、装置、计算机设备和存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本申请涉及人工智能中的机器学习技术领域,特别是涉及一种数据监控方法、装置、计算机设备和存储介质。
目前,由于大数据的发展,各种网站都需要采集用户的数据来对用户的行为进行分析。基本上都是周期性的监控来获取用户的数据并进行分析。这种较为固定的监控方式,容易浪费监控资源。例如:银行在贷款发放后,都是通过周期性的监控来及时发现用户逾期风险,比如一个月采集贷款用户数据一次,分析贷款用户逾期风险概率。发明人意识到这种较为固定的监控方式进行监控,如果监控周期较大,不利于及时发现逾期风险,数据分析效率低下,如果监控周期较为频繁,则会浪费监控资源。
发明内容
基于此,有必要针对上述技术问题,提供一种能够提高监控效率,节省监控资源的数据监控方法、装置、计算机设备和存储介质。
一种数据监控方法,所述方法包括:
接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;
根据初始监控维度在初始监控周期内获取初始账号标识对应的初始监控数据;
根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示所述未预警账号标识对应的逾期可能性大小;
将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;
根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据。
一种数据监控装置,所述装置包括:
指令接收模块,用于接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;
监控数据得到模块,用于根据初始监控维度在初始监控周期内获取初始监 控账号标识对应的初始监控数据;
概率得到模块,用于根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示未预警账号标识对应的逾期可能性大小;
模型计算模块,用于将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;
监控模块,用于根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据。
一种计算机设备,包括存储器和处理器,所述存储器存储有计算机程序,所述处理器执行所述计算机程序时实现以下步骤:
接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;
根据初始监控维度在初始监控周期内监控初始获取初始账号标识对应的初始监控数据;
根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示所述未预警账号标识对应的逾期可能性大小;
将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;
根据输出结果确定得到目标监控周期和目标监控维度,根据目标监控周期和目标监控维度获取未预警账号标识对应的目标监控数据。
一种计算机可读存储介质,其上存储有计算机程序,所述计算机程序被处理器执行时实现以下步骤:
接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;
根据初始监控维度在初始监控周期内监控初始获取初始账号标识对应的初始监控数据;
根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示所述未预警账号标识对应的逾期可能性大小;
将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;
根据输出结果确定得到目标监控周期和目标监控维度,根据目标监控周期和目标监控维度获取未预警账号标识对应的目标监控数据。
上述数据监控方法、装置、计算机设备和存储介质,通过根据初始监控数据使用评分卡模型和朴素贝叶斯模型来调整初始监控周期和初始监控维度,从 而得到目标监控周期和目标监控维度,使用目标监控周期和目标监控维度来获取目标账号标识对应的目标监控数据,实现了不同的用户使用不同的监控周期和监控维度进行监控,提高了监控的效率,节省了监控资源。
图1为一个实施例中数据监控方法的应用场景图;
图2为一个实施例中数据监控方法的流程示意图;
图3为一个实施例中训练评分卡模型的流程示意图;
图4为一个实施例中训练朴素贝叶斯模型的流程示意图;
图5为一个实施例中得到初始监控数据的流程示意图;
图6为一个实施例中得到未预警账号标识对应逾期概率的的流程示意图;
图7为一个具体实施例中可视化报表的示意图;
图8为一个实施例中数据监控装置的结构框图;
图9为一个实施例中计算机设备的内部结构图。
本申请提供的数据监控方法,可以应用于如图1所示的应用环境中。其中,终端102通过网络与服务器104进行通信。服务器104接收终端102发送的数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;服务器104根据初始监控维度在初始监控周期内监控初始获取初始账号标识对应的初始监控数据;根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示所述未预警账号标识对应的逾期可能性大小;服务器104将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;服务器104根据输出结果确定得到目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。其中,终端102可以但不限于是各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备,服务器104可以用独立的服务器或者是多个服务器组成的服务器集群来实现。
在一个实施例中,如图2所示,提供了一种数据监控方法,以该方法应用于图1中的服务器为例进行说明,包括以下步骤:
S202,接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期。
其中,初始监控账号标识用于唯一标识需要进行监控的用户,比如,在贷款平台发放贷款后要监控的贷款用户。初始监控维度是指要采集的初始监控账 号标识对应用户相关的数据字段,比如,用户常驻地址字段、用户手机话费字段和用户外部贷款申请字段等等。初始监控周期是指初步确定的采集该用户相关数据的时间长度,比如一个月,或者两个月。
具体地,服务器接收终端发送的数据监控指令,该数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期。服务器解析该数据监控指令,得到该数据监控指令携带的初始监控账号标识、初始监控维度和初始监控周期。
S204,根据初始监控维度在初始监控周期内获取初始监控账号标识对应的初始监控数据。
其中,初始监控数据是指在初始监控周期内根据初始监控维度采集初始监控账号标识对应的数据。
具体地,服务器根据初始监控维度在初始监控周期内监控初始监控账号标识,得到初始监控数据。服务器可以从互联网各个渠道获取该初始监控账号标识的初始监控维度对应的数据。也可以从内部数据库中获取到该初始监控账号标识对应的数据。
S206,根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示所述未预警账号标识对应的逾期可能性大小。
其中,未预警账号标识用于标识未符合预设预警规则的用户,该未预警账号标识包括在初始账号标识中。预设预警规则是指人为设置的一些预警规则,比如该初始监控数据中该用户的手机欠费,则该用户的账号标识为预警账号标识;该用户的常驻地址长时间无人居住,则该用户的账号标识为预警账号标识。当用户为预警账号标识需要进行风险处理,即向管理终端发送预警提示并显示预警账号标识,不用在对该预警账号标识进行监控。
评分卡模型是根据历史监控数据和评分结果使用机器学习算法训练得到的模型,该机器学习算法可以是逻辑回归算法,深度神经网络算法等等。该评分卡模型是用来预测用户的逾期可能性的模型。
具体地,服务器根据预先设置好的预警规则,从初始监控数据中确定预警账号标识和未预警账号标识,将预警账号标识和预警提示一起发送到管理终端进行展示。然后服务器将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到评分卡模型输出的评分结果,该评分结果表示该未预警账号标识对应的逾期可能性大小。
S208,将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果。
其中,朴素贝叶斯模型是预先根据历史用户贷款后的监控数据和还款是否逾期对应的评分结果使用朴素贝叶斯算法计算得到的。该输出结果可以是各个 未预警账号标识对应的预设监控维度和预设监控周期的概率值。
具体地,服务器将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中进行计算,得到各个未预警账号标识对应的预设监控维度和预设监控周期的概率值。
S210,根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据。
其中,目标监控周期是指在初始监控周期之后对未预警账号标识监控的周期,目标监控维度是指在目标监控周期中需要获取的未预警账号标识相关的数据字段。目标监控数据是指在目标监控周期内获取到的目标监控维度对应的数据。
具体地,服务器将输出结果中大于预设阈值的预设监控维度和预设监控周期作为各个未预警账号标识对应目标监控周期和目标监控维度,服务器根据所述目标监控维度在所述目标监控周期内获取各个未预警账号标识对应的目标监控数据。
在上述数据监控方法中,通过根据初始监控数据使用评分卡模型和朴素贝叶斯模型来调整初始监控周期和初始监控维度,从而得到目标监控周期和目标监控维度,使用目标监控周期和目标监控维度来获取目标账号标识对应的目标监控数据,实现了不同的用户使用不同的监控周期和监控维度进行监控,提高了监控的效率,节省了监控资源。
在一个实施例中,如图3所示,已训练的评分卡模型的生成步骤,包括:
S302,获取历史监控数据和对应的逾期结果,根据历史监控数据生成监控数据向量,根据逾期结果生成对应的评分结果向量。
其中,评分结果是指用户在贷款后是否在规定期限内还款的结果,包括逾期和未逾期。
具体地,服务器获取到历史贷款用户的监控数据和对应的逾期结果,将历史监控数据进行预处理,得到监控数据向量,即将监控数据中各个维度对应的数据数值化,然后将数值化的数据转换为向量,得到监控数据向量。将逾期结果转换为评分结果向量,比如逾期对应的评分结果向量可以为(0,1)。未预警对应的评分结果向量可以为(1,0)。
S304,将监控数据向量作为输入,将监控数据向量对应的评分结果向量作为标签使用逻辑回归算法进行训练,当达到预设条件时,得到已训练的评分卡模型。
其中,预设条件是指训练次数达到最大迭代次数或者使用的损失函数的值达到预设阈值。逻辑回归算法是指logistic回归算法,是一种广义的线性回归分 析模型。激活函数使用sigmoid函数
损失函数使用最大化似然函数。
具体地,服务器将将监控数据向量作为输入,将监控数据向量对应的评分结果作为标签使用逻辑回归算法进行训练,当训练次数达到最大迭代次数或者使用的损失函数的值达到预设阈值时,得到已训练的评分卡模型。
在上述实施例中,通过历史监控数据和对应的逾期结果使用逻辑回归算法训练得到评分卡模型,便于后续的使用,提高监控效率。
在一个实施例中,如图4所示,已训练的朴素贝叶斯模型的生成步骤,包括:
S402,获取历史监控维度、对应的监控周期和评分结果,根据历史监控维度、对应的监控周期和评分结果计算出未预警概率参数和逾期概率参数。
其中,未预警概率参数是指历史用户中贷款没有逾期的用户占所有贷款用户的概率参数,逾期概率参数是指历史用户中贷款逾期的用户占所有贷款用户的概率参数。
具体地,服务器获取到历史用户的监控维度、历史监控周期和历史评分结果。根据历史监控维度、历史监控周期和历史评分结果计算出未预警概率参数和逾期概率参数。即根据逾期的用户数和历史用户总数得到逾期概率参数。根据未逾期的用户总数和历史用户总数得到未逾期的概率参数。
S404,根据历史监控维度和评分结果计算历史监控维度的条件概率参数,根据监控周期和评分结果计算监控周期的条件概率参数,得到已训练的朴素贝叶斯模型。
具体地,服务器计算逾期结果中用户历史监控维度中各个维度的条件概率参数,即该历史监控维度中各个维度的评分结果为用户逾期的条件概率参数和历史监控维度中各个维度的评分结果为用户未逾期的条件概率参数。比如,若监控字段是手机费用字段,计算出用户贷款逾期的情况下手机欠费的条件概率和手机未欠费条件概率参数,并计算出用户贷款未逾期的情况下手机欠费的条件概率和手机未欠费条件概率参数,得到手机费用字段的条件概率参数。根据监控周期和逾期结果计算监控周期的条件概率,即在逾期的情况下不同监控周期的条件概率参数,根据计算得到的各个参数就得到了已训练的朴素贝叶斯模 型。
在上述实施例中,通过获取历史监控维度、对应的监控周期和评分结果,根据历史监控维度、对应的监控周期和评分结果计算出未预警概率参数和逾期概率参数。根据历史监控维度和评分结果计算历史监控维度的条件概率参数,根据监控周期和评分结果计算监控周期的条件概率参数,得到已训练的朴素贝叶斯模型。通过预先训练得到朴素贝叶斯模型,在进行监控时可以直接使用,方便快捷。
在一个实施例中,如图5所示,步骤S204,即根据初始监控维度在初始监控周期内获取初始监控账号标识对应的初始监控数据,包括步骤:
S502,根据初始监控维度在初始监控周期内从各个预设数据源获取监控数据报文。
其中,预设数据源是指可以获取到初始账号标识对应的各种数据的服务器。比如,从电信服务器数据库中获取用户手机费用情况,从公安服务器数据库中获取用户是否在常驻地址。从其他金融机构服务器数据库中获取该用户的贷款申请情况等等。
具体地,服务器根据初始监控维度在初始监控周期内从各个预设数据源获取各个初始监控账号标识对应的监控数据报文。从各个预设数据源获取数据时,需要预先获取到各个数据源的授权。
S504,解析各个预设数据源对应的初始监控数据报文得到解析结果。
具体地,服务器解析各个预设数据源对应的初始监控数据报文,得到解析结果。即得到各个预设数据源中对应的监控数据报文对应的数据。每个数据源都有自己的报文格式,解析不同数据源的报文得到的不同的字段数据。
S506,将解析结果转换为目标字段对应的数据,得到初始监控数据。
其中,目标字段是指预先设置好的统一字段。目标字段用于将各个数据源的报文字段统一,以便于得到相同格式的数据。
具体地,服务器按照预先设置好的字段将解析结果转换为设置好的字段对应的数据,就得到了初始监控数据。
在上述实施例中,通过将各个预设数据源的数据报文的解析结果转换为统一字段的数据,得到初始监控数据,使得到的初始监控数据的字段统一,方便后续管理。
在一个实施例中,如图6所示,在步骤S210之后,即在根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据之后,还包括步骤:
S602,根据目标监控数据得到目标监控数据向量。
具体地,服务器将该目标监控数据预处理即转换为数值化数据,然后根据数值化的目标监控数据得到目标监控数据向量。
S604,将目标监控数据向量输入到已训练的评分卡模型中,得到未预警账号标识在目标监控周期内的评分结果,根据评分结果向监控终端发送报警提示。
具体地,将目标监控数据向量输入到已训练的评分卡模型中,得到未预警账号标识在目标监控周期内的评分结果。将评分结果为逾期结果对应的未预警账号标识标记作为风险账号,向监控终端发送该风险用户对应的评分结果并进行预警提示。
在上述实施例中,根据目标监控数据得到目标监控数据向量,将目标监控数据向量输入到已训练的评分卡模型中,得到未预警账号标识在目标监控周期内的评分结果,根据评分结果向监控终端发送报警提示,通过根据监控得到的数据使用评分卡模型得到评分,能够及时提醒监控终端对评分结果中的逾期账号标识进行风险管控,避免贷款账号逾期还款。
在一个实施例中,在步骤S210之后,在根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据之后,还包括步骤:
根据目标监控数据进行统计计算,得到计算结果,根据计算结果生成可视化报表,将可视化报表发送到监控终端进行显示。
具体地,根据目标监控数据进行统计计算,得到计算结果,比如,根据目标监控数据计算月预警总人数增长率,计算人均新增申请贷款次数,计算预警类型趋势,计算风险预警人数趋势等等。然后根据计算结果生成可视化报表,将可视化报表发送到监控终端进行显示。比如,生成的报表可以如图7所示。
在上述实施例中,通过根据目标监控数据进行统计计算,得到计算结果,根据计算结果生成可视化报表,将可视化报表发送到监控终端进行显示,可以对相关监控数据进行可视化显示,方便用户查看。
应该理解的是,虽然图2-6的流程图中的各个步骤按照箭头的指示依次显示,但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明,这些步骤的执行并没有严格的顺序限制,这些步骤可以以其它的顺序执行。而且,图2-6中的至少一部分步骤可以包括多个子步骤或者多个阶段,这些子步骤或者阶段并不必然是在同一时刻执行完成,而是可以在不同的时刻执行,这些子步骤或者阶段的执行顺序也不必然是依次进行,而是可以与其它步骤或者其它步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
在一个具体的实施例中,该数据监控方法应用在贷后监控预警平台中。监控人员通过将贷后客户表上传到贷后监控预警平台中,在该贷后监控预警平台选择需要监控的贷后客户表中的客户标识和对应的初始监控字段和初始监控周期,贷后监控预警平台通过前海征信或者人行征信查询等获取到初始监控字段对应的外部字段数据,通过字段管理将外部字段转换为内部字段,然后将字段对应的数据通过规则管理进行初次预警处理,得到未预警客户标识,将未预警 客户标识对应的初始监控数据输入到评分卡模型中,得到未预警客户标识对应的贷款逾期概率,根据贷款逾期概率得到逾期结果,即将贷款逾期概率大于预设阈值的未预警客户标识标记为逾期。将该未预警客户对应的逾期概率和未预警客户对应的监控数据输入到朴素贝叶斯模型中计算得到预设监控维度和预设监控周期的概率,根据得到的概率进行初始监控字段和初始监控周期调整,即将该预设监控维度的概率大于预设阈值的预设监控维度作为目标监控维度,将该预设监控周期的概率最大的预设监控周期作为目标监控周期进行监控,并得到目标监控数据。贷后监控预警平台可以根据该目标监控数据进行统计计算,得到展示报表,将展示报表在监控终端进行显示。也可以将展示报表发送到风控人员对应的终端进行显示,便于风险控制人员进行风险处置。
在一个实施例中,如图8所示,提供了一种数据监控装置800,包括:指令接收模块802、监控数据得到模块804、概率得到模块806、模型计算模块808和监控模块810,其中:
指令接收模块802,用于接收数据监控指令,数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;
监控数据得到模块804,用于根据初始监控维度在初始监控周期内获取初始监控账号标识对应的初始监控数据;
概率得到模块806,用于根据初始监控数据确定未预警账号标识,将未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到未预警账号标识对应的评分结果,评分结果用于表示未预警账号标识对应的逾期可能性大小;
模型计算模块808,用于将未预警账号标识对应的初始监控数据和评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;
监控模块810,用于根据输出结果确定目标监控周期和目标监控维度,根据目标监控维度在目标监控周期内获取未预警账号标识对应的目标监控数据。
在一个实施例中,数据监控装置800,还包括:
向量生成模块,用于获取历史监控数据和对应的逾期结果,根据历史监控数据生成监控数据向量,根据逾期结果生成对应的评分结果向量;
模型训练模块,用于将监控数据向量作为输入,将监控数据向量对应的评分结果向量作为标签使用逻辑回归算法进行训练,当达到预设条件时,得到已训练的评分卡模型。
在一个实施例中,数据监控装置800,还包括:
参数计算模块,用于获取历史监控维度、对应的监控周期和评分结果,根据历史监控维度、对应的监控周期和评分结果计算出未预警概率参数和逾期概率参数;
条件参数计算模块,用于根据历史监控维度和评分结果计算历史监控维度 的条件概率参数,根据监控周期和评分结果计算监控周期的条件概率参数,得到已训练的朴素贝叶斯模型。
在一个实施例中,监控数据得到模块804,包括:
数据报文获取模块,用于根据初始监控维度在初始监控周期内,从各个预设数据源获取初始监控数据报文;
解析模块,用于解析各个预设数据源对应的初始监控数据报文得到解析结果;
数据转换模块,用于将解析结果转换为目标字段对应的数据,得到初始监控数据。
在一个实施例中,数据监控装置800,还包括:
目标向量得到模块,用于根据目标监控数据得到目标监控数据向量;
监控模块,用于将目标监控数据向量输入到已训练的评分卡模型中,得到未预警账号标识在目标监控周期内的评分结果,根据评分结果向监控终端发送报警提示。
在一个实施例中,数据监控装置800,还包括:
可视化生成模块,用于根据目标监控数据进行统计计算,得到计算结果,根据计算结果生成可视化报表,将可视化报表发送到监控终端进行显示。
关于数据监控装置的具体限定可以参见上文中对于数据监控方法的限定,在此不再赘述。上述数据监控装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中,也可以以软件形式存储于计算机设备中的存储器中,以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中,提供了一种计算机设备,该计算机设备可以是服务器,其内部结构图可以如图9所示。该计算机设备包括通过***总线连接的处理器、存储器、网络接口和数据库。其中,该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作***、计算机程序和数据库。该内存储器为非易失性存储介质中的操作***和计算机程序的运行提供环境。该计算机设备的数据库用于存储监控相关数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种数据监控方法,该方法包括以下步骤:接收数据监控指令,所述数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据;根据所述初始监控数据确定未预警账号标识,将所述未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到所述未预警账号标识对应的评分结果,所述评分结果用于表示所述未预警账号标识对应的逾期可能性大小;将所述未预警账号标识对应 的初始监控数据和所述评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。
本领域技术人员可以理解,图9中示出的结构,仅仅是与本申请方案相关的部分结构的框图,并不构成对本申请方案所应用于其上的计算机设备的限定,具体的计算机设备可以包括比图中所示更多或更少的部件,或者组合某些部件,或者具有不同的部件布置。
在一个实施例中,提供了一种计算机设备,包括存储器和处理器,该存储器存储有计算机程序,该处理器执行计算机程序时实现上述任意实施例中电子***生成方法所述的步骤。
在一个实施例中,提供了一种计算机可读存储介质,该存储介质为易失性存储介质或非易失性存储介质,其上存储有计算机程序,计算机程序被处理器执行时实现上述任意实施例中电子***生成方法所述的步骤,包括:接收数据监控指令,所述数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据;根据所述初始监控数据确定未预警账号标识,将所述未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到所述未预警账号标识对应的评分结果,所述评分结果用于表示所述未预警账号标识对应的逾期可能性大小;将所述未预警账号标识对应的初始监控数据和所述评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储于一非易失性计算机可读取存储介质中,该计算机程序在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM以多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
Claims (20)
- 一种数据监控方法,其中,所述方法包括:接收数据监控指令,所述数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据;根据所述初始监控数据确定未预警账号标识,将所述未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到所述未预警账号标识对应的评分结果,所述评分结果用于表示所述未预警账号标识对应的逾期可能性大小;将所述未预警账号标识对应的初始监控数据和所述评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。
- 根据权利要求1所述的方法,其中,所述已训练的评分卡模型的生成步骤,包括:获取历史监控数据和对应的逾期结果,根据所述历史监控数据生成监控数据向量,根据所述逾期结果生成评分结果向量;将所述监控数据向量作为输入,将所述监控数据向量对应的评分结果向量作为标签使用逻辑回归算法进行训练,当达到预设条件时,得到所述已训练的评分卡模型。
- 根据权利要求1所述的方法,其中,所述已训练的朴素贝叶斯模型的生成步骤,包括:获取历史监控维度、对应的监控周期和评分结果,根据所述历史监控维度、对应的监控周期和评分结果计算出未预警概率参数和逾期概率参数;根据所述历史监控维度和所述评分结果计算所述历史监控维度的条件概率参数,根据所述监控周期和所述评分结果计算所述监控周期的条件概率参数,得到所述已训练的朴素贝叶斯模型。
- 根据权利要求1所述的方法,其中,根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据,包括:根据所述初始监控维度在所述初始监控周期内从各个预设数据源获取初始监控数据报文;解析所述各个预设数据源对应的初始监控数据报文得到解析结果;将所述解析结果转换为目标字段对应的数据,得到所述初始监控数据。
- 根据权利要求1所述的方法,其中,在所述根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据之后,还包括:根据所述目标监控数据确定目标监控数据向量;将所述目标监控数据向量输入到所述已训练的评分卡模型中,得到所述未预警账号标识在所述目标监控周期内的评分结果,根据所述评分结果向监控终端发送报警提示。
- 根据权利要求1所述的方法,其中,所述根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据之后,还包括:根据所述目标监控数据进行统计计算,得到计算结果,根据所述计算结果生成可视化报表,将所述可视化报表发送到监控终端进行显示。
- 一种数据监控装置,其中,所述装置包括:指令接收模块,用于接收数据监控指令,所述数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;监控数据得到模块,用于根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据;评分计算模块,用于根据所述初始监控数据确定未预警账号标识,将所述未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到所述未预警账号标识对应的评分结果,所述评分结果用于表示所述未预警账号标识对应的逾期可能性大小;结果计算模块,用于将所述未预警账号标识对应的初始监控数据和所述评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;监控模块,用于根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。
- 根据权利要求7所述的装置,其中,所述装置还包括:向量生成模块,用于获取历史监控数据和对应的逾期结果,根据所述历史监控数据生成监控数据向量,根据所述逾期结果生成对应的评分结果向量;模型训练模块,用于将所述监控数据向量作为输入,将所述监控数据向量对应的评分结果向量作为标签使用逻辑回归算法进行训练,当达到预设条件时,得到所述已训练的评分卡模型。
- 一种计算机设备,包括存储器和处理器,所述存储器存储有计算机程序,其中,所述处理器用于执行一种数据监控方法,所述数据监控方法包括以下步骤:接收数据监控指令,所述数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据;根据所述初始监控数据确定未预警账号标识,将所述未预警账号标识对应的初始监控数据输入到已训练的评分 卡模型中,得到所述未预警账号标识对应的评分结果,所述评分结果用于表示所述未预警账号标识对应的逾期可能性大小;将所述未预警账号标识对应的初始监控数据和所述评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。
- 根据权利要求9所述的计算机设备,其中,所述已训练的评分卡模型的生成步骤,包括:获取历史监控数据和对应的逾期结果,根据所述历史监控数据生成监控数据向量,根据所述逾期结果生成评分结果向量;将所述监控数据向量作为输入,将所述监控数据向量对应的评分结果向量作为标签使用逻辑回归算法进行训练,当达到预设条件时,得到所述已训练的评分卡模型。
- 根据权利要求9所述的计算机设备,其中,所述已训练的朴素贝叶斯模型的生成步骤,包括:获取历史监控维度、对应的监控周期和评分结果,根据所述历史监控维度、对应的监控周期和评分结果计算出未预警概率参数和逾期概率参数;根据所述历史监控维度和所述评分结果计算所述历史监控维度的条件概率参数,根据所述监控周期和所述评分结果计算所述监控周期的条件概率参数,得到所述已训练的朴素贝叶斯模型。
- 根据权利要求9所述的计算机设备,其中,根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据,包括:根据所述初始监控维度在所述初始监控周期内从各个预设数据源获取初始监控数据报文;解析所述各个预设数据源对应的初始监控数据报文得到解析结果;将所述解析结果转换为目标字段对应的数据,得到所述初始监控数据。
- 根据权利要求9所述的计算机设备,其中,在所述根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据之后,还包括:根据所述目标监控数据确定目标监控数据向量;将所述目标监控数据向量输入到所述已训练的评分卡模型中,得到所述未预警账号标识在所述目标监控周期内的评分结果,根据所述评分结果向监控终端发送报警提示。
- 根据权利要求9所述的计算机设备,其中,所述根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据之后,还包括:根据所述目标监控数据进行统计计算,得到计算结果,根据所述计算结果生成可视化报表,将所述可视化报表发送到监控终端进行显示。
- 一种计算机可读存储介质,其上存储有计算机程序,其中,所述计算机程序被处理器执行时实现一种数据监控方法,所述数据监控方法包括以下步 骤:接收数据监控指令,所述数据监控指令中携带有初始监控账号标识、初始监控维度和初始监控周期;根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据;根据所述初始监控数据确定未预警账号标识,将所述未预警账号标识对应的初始监控数据输入到已训练的评分卡模型中,得到所述未预警账号标识对应的评分结果,所述评分结果用于表示所述未预警账号标识对应的逾期可能性大小;将所述未预警账号标识对应的初始监控数据和所述评分结果输入到已训练的朴素贝叶斯模型中,得到输出结果;根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据。
- 根据权利要求15所述的计算机可读存储介质,其中,所述已训练的评分卡模型的生成步骤,包括:获取历史监控数据和对应的逾期结果,根据所述历史监控数据生成监控数据向量,根据所述逾期结果生成评分结果向量;将所述监控数据向量作为输入,将所述监控数据向量对应的评分结果向量作为标签使用逻辑回归算法进行训练,当达到预设条件时,得到所述已训练的评分卡模型。
- 根据权利要求15所述的计算机可读存储介质,其中,所述已训练的朴素贝叶斯模型的生成步骤,包括:获取历史监控维度、对应的监控周期和评分结果,根据所述历史监控维度、对应的监控周期和评分结果计算出未预警概率参数和逾期概率参数;根据所述历史监控维度和所述评分结果计算所述历史监控维度的条件概率参数,根据所述监控周期和所述评分结果计算所述监控周期的条件概率参数,得到所述已训练的朴素贝叶斯模型。
- 根据权利要求15所述的计算机可读存储介质,其中,根据所述初始监控维度在所述初始监控周期内获取所述初始监控账号标识对应的初始监控数据,包括:根据所述初始监控维度在所述初始监控周期内从各个预设数据源获取初始监控数据报文;解析所述各个预设数据源对应的初始监控数据报文得到解析结果;将所述解析结果转换为目标字段对应的数据,得到所述初始监控数据。
- 根据权利要求15所述的计算机可读存储介质,其中,在所述根据所述输出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据之后,还包括:根据所述目标监控数据确定目标监控数据向量;将所述目标监控数据向量输入到所述已训练的评分卡模型中,得到所述未预警账号标识在所述目标监控周期内的评分结果,根据所述评分结果向监控终端发送报警提示。
- 根据权利要求15所述的计算机可读存储介质,其中,所述根据所述输 出结果确定目标监控周期和目标监控维度,根据所述目标监控维度在所述目标监控周期内获取所述未预警账号标识对应的目标监控数据之后,还包括:根据所述目标监控数据进行统计计算,得到计算结果,根据所述计算结果生成可视化报表,将所述可视化报表发送到监控终端进行显示。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910521514.4 | 2019-06-17 | ||
CN201910521514.4A CN110378781A (zh) | 2019-06-17 | 2019-06-17 | 数据监控方法、装置、计算机设备和存储介质 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020253381A1 true WO2020253381A1 (zh) | 2020-12-24 |
Family
ID=68249307
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/086914 WO2020253381A1 (zh) | 2019-06-17 | 2020-04-26 | 数据监控方法、装置、计算机设备和存储介质 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110378781A (zh) |
WO (1) | WO2020253381A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114694285A (zh) * | 2022-03-29 | 2022-07-01 | 重庆紫光华山智安科技有限公司 | 人流量告警方法、装置、电子设备和存储介质 |
CN115858309A (zh) * | 2022-12-30 | 2023-03-28 | 北京百度网讯科技有限公司 | 面向分布式***的数据监控方法、装置和电子设备 |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110378781A (zh) * | 2019-06-17 | 2019-10-25 | 深圳壹账通智能科技有限公司 | 数据监控方法、装置、计算机设备和存储介质 |
CN112199348A (zh) * | 2020-08-21 | 2021-01-08 | 李苗 | 基于大数据的消防监控方法、装置、计算机设备及存储介质 |
CN113344438A (zh) * | 2021-06-29 | 2021-09-03 | 百维金科(上海)信息科技有限公司 | 对贷中行为进行监控的贷款***、监控方法、设备及介质 |
CN113902549A (zh) * | 2021-10-14 | 2022-01-07 | 中国银行股份有限公司 | 一种逾期处理方法及装置 |
CN117495174B (zh) * | 2023-11-03 | 2024-07-19 | 睿智合创(北京)科技有限公司 | 一种评分卡模型的前台数据监控方法及*** |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108898418A (zh) * | 2018-05-31 | 2018-11-27 | 康键信息技术(深圳)有限公司 | 用户账号检测方法、装置、计算机设备和存储介质 |
CN109191282A (zh) * | 2018-08-23 | 2019-01-11 | 北京玖富普惠信息技术有限公司 | 一种基于行为模型的贷中监测评分方法以及*** |
CN109377333A (zh) * | 2018-09-03 | 2019-02-22 | 平安科技(深圳)有限公司 | 电子装置、基于分类模型的催收员确定方法及存储介质 |
CN109614982A (zh) * | 2018-10-18 | 2019-04-12 | 平安科技(深圳)有限公司 | 产品分析方法、装置、计算机设备以及存储介质 |
CN109829628A (zh) * | 2019-01-07 | 2019-05-31 | 平安科技(深圳)有限公司 | 基于大数据的风险预警方法、装置和计算机设备 |
CN110378781A (zh) * | 2019-06-17 | 2019-10-25 | 深圳壹账通智能科技有限公司 | 数据监控方法、装置、计算机设备和存储介质 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6119103A (en) * | 1997-05-27 | 2000-09-12 | Visa International Service Association | Financial risk prediction systems and methods therefor |
CN106651573A (zh) * | 2016-12-30 | 2017-05-10 | 中国建设银行股份有限公司 | 一种业务数据处理方法及装置 |
CN108769026B (zh) * | 2018-05-31 | 2022-02-15 | 康键信息技术(深圳)有限公司 | 用户账号检测***和方法 |
-
2019
- 2019-06-17 CN CN201910521514.4A patent/CN110378781A/zh active Pending
-
2020
- 2020-04-26 WO PCT/CN2020/086914 patent/WO2020253381A1/zh active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108898418A (zh) * | 2018-05-31 | 2018-11-27 | 康键信息技术(深圳)有限公司 | 用户账号检测方法、装置、计算机设备和存储介质 |
CN109191282A (zh) * | 2018-08-23 | 2019-01-11 | 北京玖富普惠信息技术有限公司 | 一种基于行为模型的贷中监测评分方法以及*** |
CN109377333A (zh) * | 2018-09-03 | 2019-02-22 | 平安科技(深圳)有限公司 | 电子装置、基于分类模型的催收员确定方法及存储介质 |
CN109614982A (zh) * | 2018-10-18 | 2019-04-12 | 平安科技(深圳)有限公司 | 产品分析方法、装置、计算机设备以及存储介质 |
CN109829628A (zh) * | 2019-01-07 | 2019-05-31 | 平安科技(深圳)有限公司 | 基于大数据的风险预警方法、装置和计算机设备 |
CN110378781A (zh) * | 2019-06-17 | 2019-10-25 | 深圳壹账通智能科技有限公司 | 数据监控方法、装置、计算机设备和存储介质 |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114694285A (zh) * | 2022-03-29 | 2022-07-01 | 重庆紫光华山智安科技有限公司 | 人流量告警方法、装置、电子设备和存储介质 |
CN114694285B (zh) * | 2022-03-29 | 2023-09-01 | 重庆紫光华山智安科技有限公司 | 人流量告警方法、装置、电子设备和存储介质 |
CN115858309A (zh) * | 2022-12-30 | 2023-03-28 | 北京百度网讯科技有限公司 | 面向分布式***的数据监控方法、装置和电子设备 |
CN115858309B (zh) * | 2022-12-30 | 2023-12-22 | 北京百度网讯科技有限公司 | 面向分布式***的数据监控方法、装置和电子设备 |
Also Published As
Publication number | Publication date |
---|---|
CN110378781A (zh) | 2019-10-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020253381A1 (zh) | 数据监控方法、装置、计算机设备和存储介质 | |
WO2020253358A1 (zh) | 业务数据的风控分析处理方法、装置和计算机设备 | |
WO2021184571A1 (zh) | 动态表单生成方法、装置、计算机设备和存储介质 | |
WO2022267735A1 (zh) | 业务数据处理方法、装置、计算机设备和存储介质 | |
WO2021174694A1 (zh) | 基于数据中心的运维监控方法、装置、设备及存储介质 | |
US10069891B2 (en) | Channel accessible single function micro service data collection process for light analytics | |
US20180096253A1 (en) | Rare event forecasting system and method | |
WO2019080662A1 (zh) | 信息推荐方法及装置、设备 | |
US20160224453A1 (en) | Monitoring the quality of software systems | |
US20220358789A1 (en) | Sentiment analysis of content using expression recognition | |
CN109063921A (zh) | 客户风险预警的优化处理方法、装置、计算机设备和介质 | |
CN112925911B (zh) | 基于多模态数据的投诉分类方法及其相关设备 | |
CN112507141B (zh) | 调查任务生成方法、装置、计算机设备及存储介质 | |
CN110310020B (zh) | 基于数据分析的项目方案管理方法、相关装置及存储介质 | |
CN110097250B (zh) | 产品风险预测方法、装置、计算机设备和存储介质 | |
CN116843395A (zh) | 一种业务***的告警分级方法、装置、设备及存储介质 | |
CN116681045A (zh) | 报表生成方法、装置、计算机设备及存储介质 | |
CN115860835A (zh) | 基于人工智能的广告推荐方法、装置、设备及存储介质 | |
TW201828198A (zh) | 風險識別方法及裝置 | |
CN114090407A (zh) | 基于线性回归模型的接口性能预警方法及其相关设备 | |
CN113780855A (zh) | 医疗机构监管方法、装置、计算机设备及存储介质 | |
CN111222344A (zh) | 训练神经网络的方法、装置及电子设备 | |
US20230237180A1 (en) | Systems and methods for linking a screen capture to a user support session | |
CN116993218A (zh) | 基于人工智能的指标分析方法、装置、设备及存储介质 | |
CN117271486A (zh) | 基于人工智能的数据报送方法、装置、设备及存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20827656 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20827656 Country of ref document: EP Kind code of ref document: A1 |