WO2020024448A1 - 人群绩效等级识别方法、装置、存储介质及计算机设备 - Google Patents

人群绩效等级识别方法、装置、存储介质及计算机设备 Download PDF

Info

Publication number
WO2020024448A1
WO2020024448A1 PCT/CN2018/111124 CN2018111124W WO2020024448A1 WO 2020024448 A1 WO2020024448 A1 WO 2020024448A1 CN 2018111124 W CN2018111124 W CN 2018111124W WO 2020024448 A1 WO2020024448 A1 WO 2020024448A1
Authority
WO
WIPO (PCT)
Prior art keywords
crowd
performance
user
performance level
preset
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.)
Ceased
Application number
PCT/CN2018/111124
Other languages
English (en)
French (fr)
Inventor
金戈
徐亮
肖京
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2020024448A1 publication Critical patent/WO2020024448A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques

Definitions

  • the present application relates to the field of artificial intelligence technology, and in particular, to a method, a device, a storage medium, and a computer device for identifying a performance level of a crowd.
  • the crowd performance level is usually identified only by a decision tree model, that is, the crowd performance level obtained by the decision tree model to identify the user is determined as the crowd performance level of the user.
  • the decision tree model may not be able to achieve an ideal fit.
  • the lower population is usually the target population.
  • the data of high-performance groups is usually much smaller than the data of ordinary performance groups.
  • the proportion of high-performance groups is usually 20%
  • the general-performance group is usually 80%
  • the high-performance groups are usually For the target population. If the performance level of the crowd is identified only by the decision tree model, the recognition error of the decision tree model will occur, resulting in a lower number of identified target populations and a lower recall rate of the target populations.
  • the application provides a method, device, storage medium and computer equipment for identifying the performance level of a crowd, mainly to overcome the influence of data distribution balance to a certain extent, reduce the situation of recognition error of a preset decision tree model, and ensure recognition Under the premise of accuracy, the recall rate of the target population is relatively low.
  • a method for identifying a crowd performance level including:
  • the crowd performance level corresponding to the user is determined according to the probability value of the performance level of each crowd.
  • a crowd performance level identification device including:
  • a recognition unit configured to input the performance characteristics of the crowd corresponding to the user to be identified into a preset decision tree model for recognition once, and obtain a probability value of the user belonging to the performance level of the corresponding classified crowd;
  • the determining unit is configured to determine a crowd performance level corresponding to the user according to a probability value of the performance rank of each crowd.
  • a computer non-volatile readable storage medium on which computer-readable instructions are stored, and when the computer-readable instructions are executed by a processor, the following steps are implemented:
  • the crowd performance level corresponding to the user is determined according to the probability value of the performance level of each crowd.
  • a computer device including a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor.
  • the processor executes the computer-readable instructions, Implement the following steps:
  • the crowd performance level corresponding to the user is determined according to the probability value of the performance level of each crowd.
  • a crowd performance level recognition method, device, storage medium and computer equipment provided by the present application.
  • the present application can input a crowd performance characteristic corresponding to a user to be identified into a preset
  • the decision tree model performs one identification to obtain the probability value of the user belonging to the performance level of the corresponding classified population.
  • the performance characteristics of the crowd and the probability value of the user belonging to the corresponding classified population performance level can be input to a preset secondary recognition model for secondary recognition to obtain the probability that the user belongs to the performance level of each crowd.
  • a preset secondary recognition model for secondary recognition to obtain the probability that the user belongs to the performance level of each crowd.
  • the user can use the preset secondary recognition model to perform secondary recognition on the recognition result of the preset decision tree model, which can overcome the influence of data distribution balance to a certain extent, and reduce the recognition error of the preset decision tree model. In the case of ensuring the accuracy of recognition, it can improve the recall rate of the target population that is relatively low.
  • FIG. 1 shows a flowchart of a method for identifying a crowd performance level according to an embodiment of the present application
  • FIG. 2 shows a flowchart of another method for identifying a performance level of a crowd according to an embodiment of the present application
  • FIG. 3 is a schematic structural diagram of a crowd performance level identification device according to an embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of another device for identifying a performance level of a crowd according to an embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of a computer device according to an embodiment of the present application.
  • the crowd performance level is usually identified only by a decision tree model, that is, the crowd performance level obtained by the decision tree model to identify the user is determined as the crowd performance level of the user.
  • the decision tree model may not be able to achieve an ideal fit.
  • the lower population is usually the target population. If the performance level of the crowd is identified only by the decision tree model, the recognition error of the decision tree model will occur, resulting in a lower number of identified target populations and a lower recall rate of the target populations.
  • an embodiment of the present application provides a method for identifying a performance level of a crowd. As shown in FIG. 1, the method includes:
  • the performance characteristics of the crowd may include, but are not limited to, the number of monthly average courses, the latitude and longitude of the work address, and the number of Internet trading products in a single month.
  • the crowd performance level may include a high performance level and a general performance level.
  • the preset decision tree model may be established according to a sample crowd performance characteristic and a sample crowd performance level, and the preset decision tree model may save a correspondence rule between a crowd performance characteristic and a crowd performance level.
  • the classified crowd performance level corresponding to the user may be a crowd performance level obtained by classifying the user according to the preset decision tree model, and may specifically be obtained by classifying according to input crowd performance characteristics, that is, each crowd performance characteristic Both can be used as nodes to identify the user category. After the crowd performance characteristics are input into the preset decision tree model, the preset decision tree model will classify the user according to the crowd performance characteristics, and finally obtain the crowd performance level of the user.
  • the average number of monthly learning courses is 50
  • the work address latitude and longitude are (123.435, 41.819)
  • the number of Internet trading products in a single month is 10.
  • the preset decision tree model first classifies users based on the average number of monthly learning courses 50 , And then classify users according to the latitude and longitude of the work address (123.435, 41.819); classify users according to the number of Internet transactions in a single month, 10, and finally get the performance level of the classified crowd corresponding to the user, and calculate the user attribution The probability value corresponding to the performance level of the corresponding classified population.
  • the performance characteristics of the corresponding population of the user are: the average number of courses per month is 50, the latitude and longitude of the work address is (123.435, 41.819), and the number of Internet trading products in a single month is 10; the preset decision tree model identifies the user
  • the corresponding classified population performance level is high performance level, and the probability value of the user belonging to the high performance level is calculated to be 0.7, then the average number of monthly learning courses is 50, the work address latitude and longitude (123.435, 41.819), and the Internet in a single month.
  • the number of traded products 10, the probability value 0.7 belonging to the high performance level is input as a feature to a preset logistic regression model to identify the user twice, and the probability value of the user belonging to the high performance level and the user belonging to the General performance level.
  • the crowd performance level corresponding to the highest probability value may be determined as the crowd performance level corresponding to the user.
  • the high performance level may be determined as the performance level of the crowd corresponding to the user, that is, the user is determined to be a high performance crowd.
  • a method for identifying a crowd performance level provided in the embodiment of the present application.
  • the embodiment of the present application can input a crowd performance characteristic corresponding to a user to be identified into a preset decision tree model. Once identified, the probability value of the user belonging to the performance level of the corresponding classified population is obtained.
  • the performance characteristics of the crowd and the probability value of the user belonging to the corresponding classified population performance level can be input to a preset secondary recognition model for secondary recognition to obtain the probability that the user belongs to the performance level of each crowd.
  • this embodiment of the present application provides another method for identifying the performance level of the crowd, as shown in FIG. 2.
  • Methods include:
  • the crowd performance data may be manually uploaded, or may be collected from an enterprise performance management system.
  • the crowd performance characteristics corresponding to the user may be extracted from the crowd performance data by way of feature keyword matching. Specifically, the keywords of the crowd performance characteristics can be matched with the crowd performance data, so as to extract the crowd performance characteristics corresponding to the user from the crowd performance data.
  • the crowd performance characteristic keyword is “average learning courses per month”, the number of monthly average learning courses corresponding to the user may be extracted from the crowd performance data, which may specifically be 80; if the crowd performance characteristic keyword is "Working address latitude and longitude", the working address longitude and latitude corresponding to the user can be extracted from the crowd performance data, which can be specifically (123.436, 41.819). If the crowd performance characteristic keyword is "Internet trading products in a single month”, the number of Internet trading products corresponding to the user in a single month may be extracted from the crowd performance data, which may specifically be 20.
  • the method before step 201 may further include: obtaining sample population performance characteristics and sample population performance levels corresponding to multiple sample users; according to the sample population
  • the performance characteristics and the performance level of the sample population are used to establish the preset decision tree model.
  • the sample population performance characteristics and the sample population performance level can be trained by a decision tree algorithm to obtain the preset decision tree model.
  • the preset decision tree model stores correspondence rules between crowd performance characteristics and crowd performance levels.
  • Decision tree algorithm can be a method to approximate the value of discrete functions. It is a typical classification method. It first processes the data, uses inductive algorithms to generate readable rules and decision trees, and then uses the decisions to analyze new data.
  • a decision tree is essentially a process of classifying data through a series of rules.
  • the method may further include: obtaining a probability value of the plurality of sample users calculated by the preset decision tree model belonging to the performance level of the corresponding classified population; according to the sample population
  • the performance characteristic and the probability value of the performance level of each sample population are used to establish the preset secondary recognition model.
  • the sample population performance characteristics can be input to the preset decision tree model for identification, and the probability values of the multiple sample users belonging to the performance levels of the various populations can be obtained.
  • a preset logistic regression algorithm may be used to analyze the performance characteristics of the sample population and the plurality of sample users at the performance level of each population Training with a probability value of 50 to establish the preset logistic regression model.
  • the preset logistic regression algorithm also called logistic regression analysis algorithm, is one of the classification and prediction algorithms. The probability of future outcomes can be predicted by the performance of historical data.
  • the logistic regression algorithm trains the performance characteristics of the sample population corresponding to each sample user and the probability values of the plurality of sample users belonging to the performance level of the corresponding classified population
  • the performance characteristics of the sample population and The probability value of the plurality of sample users belonging to the performance level of the corresponding classified population is used as an independent variable
  • the performance level of the sample population is used as a dependent variable to train the preset logistic regression model.
  • the preset secondary recognition model is a preset neural network model
  • using the preset neural network algorithm to train the sample population performance characteristics and the probability values of the plurality of sample users belonging to the performance levels of each population Establishing the preset neural network model.
  • the preset neural network algorithm may be a convolutional neural network algorithm, a recursive neural network algorithm, and the like, and the preset neural network model may include an input layer, a hidden layer, and an output layer.
  • the input layer may correspond to a training set consisting of the input sample population performance characteristics and the probability values of the plurality of sample users belonging to the corresponding classified population performance level
  • the hidden layer may include a feature extraction layer and a calculation layer
  • the feature extraction A layer can be used to extract the performance characteristics of the sample population and the probability values of the plurality of sample users belonging to the corresponding classification population performance level according to the training set
  • the calculation layer can be used to calculate the probability value calculation of the performance level of each population
  • the output layer is used to output the calculated probability values of the performance levels of each group of people.
  • step 101 The process of obtaining the probability value of the user belonging to the performance level of the corresponding classified group has been described in detail in step 101, which is not described in the embodiment of the present application.
  • the preset secondary recognition model may be a preset logistic regression model or a preset neural network model.
  • the probability value that the user belongs to the performance level of the corresponding classified crowd may be a separate feature similar to the crowd performance feature. That is, the crowd performance feature and the probability value of the user belonging to the corresponding classified crowd performance level may be input as common features to a preset secondary recognition model to perform secondary recognition on the user.
  • the preset secondary recognition model is a preset logistic regression model
  • the probability values of the performance level of the user belonging to each population can be calculated by the following formula:
  • x (x 1 , x 2 , ... x p ), x 1 , x 2 , ..., x p-1 can be the performance characteristics of the crowd, and x p can be the user belonging to the corresponding classified population
  • the probability value of the performance level (as a separate feature). Therefore, the preset secondary recognition model may use the crowd performance characteristics and the probability value of the user belonging to the corresponding classified crowd performance level as input features or a training set to calculate, to obtain the user belonging to each crowd performance level Probability value.
  • the user performs secondary identification, so the output of the preset logistic regression model can be obtained: the probability that the user belongs to the high performance level is 0.83, and that the user belongs to the ordinary performance level is 0.17.
  • the preset secondary recognition model calculates the probability that a user belongs to a high performance level is 0.56, the probability that a user belongs to a normal performance level is 0.46, and the user has the highest probability value that belongs to a high performance level, then
  • the high performance level is determined as the performance level of the crowd corresponding to the user, that is, the user is determined as the high performance crowd.
  • the preset secondary recognition model calculates that the probability that a user belongs to a high performance level is 0.32, calculates that the probability that a user belongs to a normal performance level is 0.68, and the user has the highest probability that the user belongs to a normal performance level, the general The performance level is determined as the performance level of the crowd corresponding to the user, that is, the user is determined to be a general performance crowd.
  • the embodiment of the present application can input the performance characteristics of the crowd corresponding to the user to be identified into a preset decision tree model The identification is performed once to obtain the probability value of the user belonging to the performance level of the corresponding classified population.
  • the performance characteristics of the crowd and the probability value of the user belonging to the corresponding classified population performance level can be input to a preset secondary recognition model for secondary recognition to obtain the probability that the user belongs to the performance level of each crowd. And can determine the performance level of the crowd corresponding to the user according to the probability value.
  • the user can use the preset secondary recognition model to perform secondary recognition on the recognition result of the preset decision tree model, which can overcome the influence of data distribution balance to a certain extent, and reduce the recognition error of the preset decision tree model. In the case of ensuring the accuracy of recognition, it can improve the recall rate of the target population that is relatively low.
  • an embodiment of the present application provides a device for identifying a crowd performance level.
  • the device includes a recognition unit 31 and a determination unit 32.
  • the identification unit 31 may be configured to input the performance characteristics of the crowd corresponding to the user to be identified into a preset decision tree model for identification once, and obtain a probability value that the user belongs to the performance level of the corresponding classified crowd.
  • the identification unit 31 is a main functional module in this device that inputs the performance characteristics of the crowd corresponding to the user to be identified into a preset decision tree model for identification, and obtains the probability value that the user belongs to the performance level of the corresponding classified crowd.
  • the identification unit 31 may be further configured to input the crowd performance characteristics and the probability value of the user belonging to the corresponding classified crowd performance level into a preset secondary recognition model for secondary recognition, to obtain that the user belongs to each The probability value of the crowd performance level; the recognition unit 31 also inputs the crowd performance characteristics and the probability value of the user belonging to the corresponding category of the crowd performance level into the preset secondary recognition model for secondary recognition in the device.
  • the main functional module of the probability value of the user belonging to the performance level of each group is also the core module.
  • the determining unit 32 may be configured to determine a crowd performance level corresponding to the user according to a probability value of the performance rank of each crowd.
  • the determining unit 32 is a main functional module of the device that determines the performance level of the crowd corresponding to the user according to the probability value of the performance level of each crowd.
  • the determining unit 31 may be specifically configured to sort the probability values of the performance levels of the respective crowds; and the performance level of the crowd corresponding to the highest probability value To determine the performance level of the crowd corresponding to the user.
  • the apparatus in order to establish the preset decision tree model and the preset secondary recognition model, the apparatus further includes: an obtaining unit 33 and a establishing unit 34, as shown in FIG. 4.
  • the obtaining unit 33 may be configured to obtain a sample population performance characteristic and a sample population performance level corresponding to a plurality of sample users.
  • the obtaining unit 34 is a main functional module for obtaining the sample population performance characteristics and the sample population performance level corresponding to a plurality of sample users in the device.
  • the establishing unit 34 may be configured to establish the preset decision tree model according to the sample population performance characteristics and the sample population performance level.
  • the establishing unit 34 is a main function module of the device for establishing the preset decision tree model according to the performance characteristics of the sample population and the performance level of the sample population.
  • the obtaining unit 33 may be further configured to obtain a probability value of the plurality of sample users calculated by the preset decision tree model belonging to the performance level of the corresponding classified population.
  • the obtaining unit 34 is also a main functional module in the device that obtains the probability values of the plurality of sample users calculated by the preset decision tree model belonging to the performance level of the corresponding classified population.
  • the establishing unit 34 may be further configured to establish the preset secondary recognition model according to the performance characteristics of the sample population and the probability value of the performance level of each sample population.
  • the establishing unit 34 is also a main functional module in the device for establishing the preset secondary recognition model according to the performance characteristics of the sample population and the probability value of the performance level of each sample population.
  • the establishing unit 34 may be specifically configured to: when the preset secondary recognition model is a preset logistic regression model, use a preset logistic regression algorithm to analyze the performance characteristics of the sample population and the multiple Probability values of the sample users belonging to the performance level of the corresponding classified group are trained to establish the preset logistic regression model.
  • the establishing unit 34 may be specifically configured to: when the preset secondary recognition model is a preset neural network model, use the preset neural network algorithm to attribute the sample population performance characteristics and the plurality of sample users to The training is performed corresponding to the probability value of the performance level of the classified crowd, and the preset neural network model is established.
  • the obtaining unit 33 may be specifically configured to input the performance characteristics of the sample population into the preset decision tree model for identification, and obtain a probability value that the user belongs to the performance level of the corresponding classified population.
  • the apparatus may further include an extraction unit 35.
  • the obtaining unit 33 may be further configured to obtain crowd performance data corresponding to the user to be identified.
  • the obtaining unit 33 is also a main functional module in the device for obtaining the performance data of the crowd corresponding to the user to be identified.
  • the extraction unit 35 may be configured to extract a crowd performance characteristic corresponding to the user from the crowd performance data.
  • the extraction unit 35 is a main functional module for extracting the crowd performance characteristics corresponding to the user from the crowd performance data in the device.
  • an embodiment of the present application further provides a computer non-volatile readable storage medium that stores computer-readable instructions.
  • the computer-readable instructions are executed by a processor, The following steps are implemented: inputting the performance characteristics of the crowd corresponding to the user to be identified into a preset decision tree model for identification once, and obtaining the probability value of the user belonging to the corresponding classification crowd performance level; the performance characteristics of the crowd and the belonging of the user
  • the probability value corresponding to the performance level of the classified crowd is input to a preset secondary recognition model for secondary recognition to obtain the probability value of the user belonging to the performance level of each crowd; according to the probability value, the performance of the crowd corresponding to the user is determined grade.
  • an embodiment of the present application further provides a physical structure diagram of a computer device.
  • the computer device includes: The processor 41, the memory 42, and computer-readable instructions stored on the memory 42 and executable on the processor, wherein the memory 42 and the processor 41 are both arranged on a bus 43 and the processor 41 executes the computer-readable
  • the following steps are implemented: inputting the performance characteristics of the crowd corresponding to the user to be identified into a preset decision tree model for identification, and obtaining the probability value of the user belonging to the corresponding classification crowd performance level;
  • the probability value of the user belonging to the performance level of the corresponding classified population is input to a preset secondary recognition model for secondary recognition, and the probability value of the user belonging to the performance level of each population is obtained; according to the probability value, the corresponding value of the user is determined.
  • Crowd Performance Level The computer device further includes a bus 43 configured to be
  • the performance characteristics of the crowd corresponding to the user to be identified can be input to a preset decision tree model for identification once, and the probability value of the user belonging to the performance level of the corresponding classified crowd can be obtained.
  • the performance characteristics of the crowd and the probability value of the user belonging to the corresponding classified population performance level can be input to a preset secondary recognition model for secondary recognition to obtain the probability that the user belongs to the performance level of each crowd. And can determine the performance level of the crowd corresponding to the user according to the probability value.
  • the user can use the preset secondary recognition model to perform secondary recognition on the recognition result of the preset decision tree model, which can overcome the influence of data distribution balance to a certain extent, and reduce the recognition error of the preset decision tree model. In the case of ensuring the accuracy of recognition, it can improve the recall rate of the target population that is relatively low.
  • modules or steps of the present application may be implemented by a general-purpose computing device, and they may be concentrated on a single computing device or distributed in a network composed of multiple computing devices.
  • they can be implemented with computer-readable instruction code executable by the computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, can be different from this
  • the steps shown or described are performed in sequence, either by making them into individual integrated circuit modules, or by making multiple modules or steps into a single integrated circuit module. As such, this application is not limited to any particular combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Data Mining & Analysis (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种人群绩效等级识别方法、装置、存储介质及计算机设备，涉及人工智能技术领域，主要目的在于能够在一定程度上克服数据分布平衡的影响，减少预设决策树模型识别错误的情况，且能够在保证识别精确度的前提下，提升占比较低的目标人群召回率。所述方法包括：将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值（101）；将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值（102）；根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级（103）。

Description

人群绩效等级识别方法、装置、存储介质及计算机设备
本申请要求与2018年8月1日提交中国专利局、申请号为2018108664329、申请名称为“人群绩效等级识别方法、装置、存储介质及计算机设备”的中国专利申请的优先权，其全部内容通过引用结合在申请中。
技术领域
本申请涉及人工智能技术领域，尤其是涉及一种人群绩效等级识别方法、装置、存储介质及计算机设备。
背景技术
近年来，很多行业开始重视人群绩效，尤其是开始重视识别人群绩效等级，通过识别出目标人群并对目标人群进行奖励，能够大大促进企业整体效益的提升。
目前，在识别人群绩效等级时，通常仅通过决策树模型识别人群绩效等级，即将决策树模型对用户识别得到的人群绩效等级，确定为用户的人群绩效等级。然而，由于人群数据量庞大且各个人群绩效等级的人群数据分布不平衡，决策树模型可能无法取得理想拟合效果，通常占比较低的人群通常为目标人群。例如，高绩效等级的人群数据通常会远小于普通绩效等级的人群数据，一般高绩效等级的人群占比通常为20％，普通绩效等级的人群占比通常为80％，高绩效等级的人群通常为目标人群。若仅通过决策树模型识别人群绩效等级，会发生决策树模型识别错误的情况，造成识别出来的占比较低的目标人群的数量较少，从而造成占比较低的目标人群召回率较低。
发明内容
本申请提供了一种人群绩效等级识别方法、装置、存储介质及计算机设备，主要在于能够在一定程度上克服数据分布平衡的影响，减少预设决策树模型识别错误的情况，且能够在保证识别精确度的前提下，提升占比较低的目标人群召回率。
根据本申请的第一个方面，提供一种人群绩效等级识别方法，包括：
将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
根据本申请的第二个方面，提供一种人群绩效等级识别装置，包括：
识别单元，用于将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
所述识别单元，还用于将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
确定单元，用于根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
根据本申请的第三个方面，提供一种计算机非易失性可读存储介质，其上存储有计算机可读指令，该计算机可读指令被处理器执行时实现以下步骤：
将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
根据本申请的第四个方面，提供一种计算机设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现以下步骤：
将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
本申请提供的一种人群绩效等级识别方法、装置、存储介质及计算机设备，与目前仅通过决策树模型识别人群绩效等级相比，本申请能够将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。与此同时，能够将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；并能够根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。从而能 够实现在预设决策树模型的识别结果上，利用预设二次识别模型对所述用户进行二次识别，能够在一定程度上克服数据分布平衡的影响，减少预设决策树模型识别错误的情况，且能够在保证识别精确度的前提下，提升占比较低的目标人群召回率。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1示出了本申请实施例提供的一种人群绩效等级识别方法流程图；
图2示出了本申请实施例提供的另一种人群绩效等级识别方法流程图；
图3示出了本申请实施例提供的一种人群绩效等级识别装置的结构示意图；
图4示出了本申请实施例提供的另一种人群绩效等级识别装置的结构示意图；
图5示出了本申请实施例提供的一种计算机设备的实体结构示意图。
具体实施方式
下文中将参考附图并结合实施例来详细说明本申请。需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。
如背景技术，目前，在识别人群绩效等级时，通常仅通过决策树模型识别人群绩效等级，即将决策树模型对用户识别得到的人群绩效等级，确定为用户的人群绩效等级。然而，由于人群数据量庞大且各个人群绩效等级的人群数据分布不平衡，决策树模型可能无法取得理想拟合效果，通常占比较低的人群通常为目标人群。若仅通过决策树模型识别人群绩效等级，会发生决策树模型识别错误的情况，造成识别出来的占比较低的目标人群的数量较少，从而造成占比较低的目标人群召回率较低。
为了解决上述问题，本申请实施例提供了一种人群绩效等级识别方法，如图1所示，所述方法包括：
101、将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。
其中，所述人群绩效特征可以包括但不限于：月均学习课程数、工作地址经纬度、单月内互联网交易产品数。所述人群绩效等级可以包括：高绩效等级和普遍绩效等级。所述预设决策树模型可以为根据样本人群绩效特征和样本人群绩效等级建立的，所述预设决策树模型可以保存有人群绩效特征和人群绩效等级之间的对应规则。所述用户对应的分类人群绩效等级可以为所述预设决策树模型对所述用户进行分类得到的人群绩效等级，具体地 可以为根据输入的人群绩效特征分类得到的，即每个人群绩效特征都可以作为识别所述用户类别的节点。在人群绩效特征输入到预设决策树模型后，预设决策树模型会根据人群绩效特征对所述用户进行分类，最后得到所述用户的人群绩效等级。
例如，月均学习课程数为50、工作地址经纬度为(123.435，41.819)、单月内互联网交易产品数为10，所述预设决策树模型会先根据月均学习课程数50对用户进行分类，然后根据工作地址经纬度(123.435，41.819)对用户进行分类；根据单月内互联网交易产品数10对用户进行分类，最后可以得到所述用户对应的分类人群绩效等级，并计算出所述用户归属于对应分类人群绩效等级的概率值。
102、将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值。
例如，若用户对应的人群绩效特征为：月均学习课程数为50、工作地址经纬度为(123.435，41.819)、单月内互联网交易产品数为10；所述预设决策树模型识别所述用户对应的分类人群绩效等级为高绩效等级，且计算所述用户归属于高绩效等级的概率值为0.7，则可以将月均学习课程数50、工作地址经纬度(123.435，41.819)、单月内互联网交易产品数10、归属于高绩效等级的概率值0.7作为特征输入到预设逻辑回归模型对所述用户进行二次识别，得到所述用户归属于高绩效等级的概率值以及所述用户归属于普通绩效等级。
103、根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
需要说明的是，可以将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。例如，用户归属于高绩效等级的概率值为所有概率值中最高的，则可以将高绩效等级确定为所述用户对应的人群绩效等级，即确定用户为高绩效人群。
本申请实施例提供的一种人群绩效等级识别方法，与目前仅通过决策树模型识别人群绩效等级相比，本申请实施例能够将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。与此同时，能够将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；并能够根据所述概率值，确定所述用户对应的人群绩效等级，从而能够实现在预设决策树模型的识别结果上，利用预设二次识别模型对所述用户进行二次识别，能够在一定程度上克服数据分布平衡的影响，减少预设决策树模型识别错误的情况，且能够在保证识别精确度的前提下，提升占比较低的目标人群召回率。
进一步的，为了更好的说明上述人群绩效等级识别的过程，作为对上述实施例的细化和扩展，本申请实施例提供了另一种人群绩效等级识别方法，如图2所示，所述方法包括：
201、获取待识别用户对应的人群绩效数据，并从所述人群绩效数据中提取出所述用户对应的人群绩效特征。
其中，所述人群绩效数据可以为人工上传的，也可以为从企业的绩效管控系统采集的。在本申请实施例中，可以通过特征关键字匹配的方式从人群绩效数据中提取出所述用户对应的人群绩效特征。具体地，可以将人群绩效特征关键字与所述人群绩效数据进行匹配，实现从所述人群绩效数据中提取出所述用户对应的人群绩效特征。
例如，若人群绩效特征关键字为“月均学习课程”，则可以从所述人群绩效数据中提取出所述用户对应的月均学习课程数，具体可以为80；若人群绩效特征关键字为“工作地址经纬度”，则可以从从所述人群绩效数据中提取出所述用户对应的工作地址经纬度，具体可以为(123.436，41.819)。若人群绩效特征关键字为“单月内互联网交易产品”，则可以从所述人群绩效数据中提取出所述用户对应的单月内互联网交易产品数，具体可以为20。
对于本申请实施例，为了建立所述预设决策树模型，所述步骤201之前所述方法还可以包括：获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级；根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型。具体地，可以通过决策树算法对所述样本人群绩效特征和所述样本人群绩效等级进行训练，得到所述预设决策树模型。所述预设决策树模型保存有人群绩效特征和人群绩效等级之间的对应规则。决策树算法可以是一种逼近离散函数值的方法。它是一种典型的分类方法，首先对数据进行处理，利用归纳算法生成可读的规则和决策树，然后使用决策对新数据进行分析。本质上决策树是通过一系列规则对数据进行分类的过程。
此外，为了建立所述二次识别模型，所述方法还可以包括：获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值；根据所述样本人群绩效特征和所述各个样本人群绩效等级的概率值，建立所述预设二次识别模型。具体地，可以通过将所述样本人群绩效特征输入到所述预设决策树模型进行识别，得到所述多个样本用户归属于各个人群绩效等级的概率值。
在具体应用场景中，若所述预设二次识别模型为预设逻辑回归模型，则可以利用预设逻辑回归算法对所述样本人群绩效特征和所述多个样本用户归属于各个人群绩效等级的概率值进行训练，建立所述预设逻辑回归模型。预设逻辑回归算法又称为逻辑回归分析算 法，是分类和预测算法中的一种。可以通过历史数据的表现对未来结果发生的概率进行预测。在本申请实施例中，逻辑回归算法对所述各个样本用户对应的样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值进行训练时，可以将样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值作为自变量，把样本人群绩效等级作为因变量训练所述预设逻辑回归模型。
或者若所述预设二次识别模型为预设神经网络模型，则利用预设神经网络算法对所述样本人群绩效特征和所述多个样本用户归属于各个人群绩效等级的概率值进行训练，建立所述预设神经网络模型。所述预设神经网络算法可以为卷积神经网络算法、递归神经网络算法等，所述预设神经网络模型可以包括输入层、隐藏层、输出层。输入层可以对应于输入的样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值组成的训练集，所述隐藏层可以包括特征提取层和计算层，所述特征提取层可以用于根据所述训练集提取样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值，所述计算层可以用于计算各个人群绩效等级的概率值计算；所述输出层用于将计算的各个人群绩效等级的概率值进行输出。
202、将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。
其中，得到所述用户归属于对应分类人群绩效等级的概率值的过程在步骤101已详细阐述，本申请实施例在此不进行赘述。
203、将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值。
其中，所述预设二次识别模型可以为预设逻辑回归模型或者预设神经网络模型。在所述预设二次识别模型中所述用户归属于对应分类人群绩效等级的概率值可以为与所述人群绩效特征类似的单独特征。即可以将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值作为共同特征输入到预设二次识别模型对所述用户进行二次识别。具体地，当所述预设二次识别模型为预设逻辑回归模型时，可以通过如下公式计算得到所述用户归属于各个人群绩效等级的概率值：
Figure PCTCN2018111124-appb-000001
其中，x＝(x 1，x 2，...x p)，x 1，x 2，...,x p-1可以为人群绩效特征、x p可以为所述用户归属于对应分类人群绩效等级的概率值(作为一个单独特征)。因此，所述预设二次识 别模型可以以所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值作为输入特征或者训练集进行计算，得到所述用户归属于各个人群绩效等级的概率值。
例如，可以将月均学习课程数80、工作地址经纬度(143.435，51.819)、单月内互联网交易产品数30、归属于高绩效等级的概率值0.6作为共同特征输入到预设逻辑回归模型对所述用户进行二次识别，因此，可以得到所述预设逻辑回归模型输出的：所述用户归属于高绩效等级的概率值为0.83，所述用户归属于普通绩效等级为0.17。
204、对所述各个人群绩效等级的概率值进行排序，并将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。
例如，若所述预设二次识别模型计算用户归属于高绩效等级的概率值为0.56，用户归属于普通绩效等级的概率值为0.46，用户归属于高绩效等级的概率值最高，则可以将高绩效等级确定为所述用户对应的人群绩效等级，即确定用户为高绩效人群。若所述预设二次识别模型计算用户归属于高绩效等级的概率值为0.32，计算用户归属于普通绩效等级的概率值为0.68，用户归属于普通绩效等级的概率值最高，则可以将普通绩效等级确定为所述用户对应的人群绩效等级，即确定用户为普通绩效人群。
本申请实施例提供的另一种人群绩效等级识别方法，与目前仅通过决策树模型识别人群绩效等级相比，本申请实施例能够将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。与此同时，能够将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；并能够根据所述概率值，确定所述用户对应的人群绩效等级。从而能够实现在预设决策树模型的识别结果上，利用预设二次识别模型对所述用户进行二次识别，能够在一定程度上克服数据分布平衡的影响，减少预设决策树模型识别错误的情况，且能够在保证识别精确度的前提下，提升占比较低的目标人群召回率。
进一步地，作为图1的具体实现，本申请实施例提供了一种人群绩效等级识别装置，如图3所示，所述装置包括：识别单元31和确定单元32。
所述识别单元31，可以用于将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。所述识别单元31是本装置中将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值的主要功能模块。
所述识别单元31，还可以用于将所述人群绩效特征和所述用户归属于对应分类人群绩 效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；所述识别单元31还是本装置中将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值的主要功能模块，也是核心模块。
所述确定单元32，可以用于根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。所述确定单元32是本装置中根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级的主要功能模块。
对于本申请实施例，为了确定为所述用户对应的人群绩效等级，所述确定单元31，具体可以用于对所述各个人群绩效等级的概率值进行排序；将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。
对于本申请实施例，为了建立所述预设决策树模型和所述预设二次识别模型，所述装置还包括：获取单元33和建立单元34，如图4所示。
所述获取单元33，可以用于获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级。所述获取单元34是本装置中获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级的主要功能模块。
所述建立单元34，可以用于根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型。所述建立单元34是本装置中根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型的主要功能模块。
所述获取单元33，还可以用于获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值。所述获取单元34还是本装置中获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值的主要功能模块。
所述建立单元34，还可以用于根据所述样本人群绩效特征和所述各个人样本群绩效等级的概率值，建立所述预设二次识别模型。所述建立单元34还是本装置中根据所述样本人群绩效特征和所述各个人样本群绩效等级的概率值，建立所述预设二次识别模型的主要功能模块。
在具体应用场景中，所述建立单元34，具体可以用于当所述预设二次识别模型为预设逻辑回归模型时，利用预设逻辑回归算法对所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值进行训练，建立所述预设逻辑回归模型。
所述建立单元34，具体还可以用于当所述预设二次识别模型为预设神经网络模型时，利用预设神经网络算法对所述样本人群绩效特征和所述多个样本用户归属于对应分类人 群绩效等级的概率值进行训练，建立所述预设神经网络模型。
所述获取单元33，具体可以用于将所述样本人群绩效特征输入到所述预设决策树模型进行识别，得到所述用户归属于对应分类人群绩效等级的概率值。
对于本申请实施例，为了获取待识别用户对应的人群绩效特征，所述装置还可以包括：提取单元35。
所述获取单元33，还可以用于获取待识别用户对应的人群绩效数据。所述获取单元33还是本装置中获取待识别用户对应的人群绩效数据的主要功能模块。
所述提取单元35，可以用于从所述人群绩效数据中提取出所述用户对应的人群绩效特征。所述提取单元35是本装置中从所述人群绩效数据中提取出所述用户对应的人群绩效特征的主要功能模块。
需要说明的是，本申请实施例提供的一种人群绩效等级识别装置所涉及各功能模块的其他相应描述，可以参考图1所示方法的对应描述，在此不再赘述。
基于上述如图1所示方法，相应的，本申请实施例还提供了一种计算机非易失性可读存储介质，其上存储有计算机可读指令，该计算机可读指令被处理器执行时实现以下步骤：将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；根据所述概率值，确定所述用户对应的人群绩效等级。
基于上述如图1所示方法和如图3所示人群绩效等级识别装置的实施例，本申请实施例还提供了一种计算机设备的实体结构图，如图5所示，该计算机设备包括：处理器41、存储器42、及存储在存储器42上并可在处理器上运行的计算机可读指令，其中存储器42和处理器41均设置在总线43上所述处理器41执行所述计算机可读指令时实现以下步骤：将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；根据所述概率值，确定所述用户对应的人群绩效等级。该计算机设备还包括：总线43，被配置为耦接处理器41及存储器42。
通过本申请的技术方案，能够将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值。与此同时，能够将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次 识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；并能够根据所述概率值，确定所述用户对应的人群绩效等级。从而能够实现在预设决策树模型的识别结果上，利用预设二次识别模型对所述用户进行二次识别，能够在一定程度上克服数据分布平衡的影响，减少预设决策树模型识别错误的情况，且能够在保证识别精确度的前提下，提升占比较低的目标人群召回率。
显然，本领域的技术人员应该明白，上述的本申请的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个的计算装置上，或者分布在多个计算装置所组成的网络上，可选地，它们可以用计算装置可执行的计算机可读指令代码来实现，从而，可以将它们存储在存储装置中由计算装置来执行，并且在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本申请不限制于任何特定的硬件和软件结合。
以上所述仅为本申请的优选实施例而已，并不用于限制本申请，对于本领域的技术人员来说，本申请可以有各种更改和变化。凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包括在本申请的保护范围之内。

Claims (20)

  1. 一种人群绩效等级识别方法，其特征在于，包括：
    将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
    将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
    根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级，包括：
    对所述各个人群绩效等级的概率值进行排序；
    将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。
  3. 根据权利要求1所述的方法，其特征在于，所述将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值之前，所述方法还包括：
    获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级；
    根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型，并获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值；
    根据所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值，建立所述预设二次识别模型。
  4. 根据权利要求3所述的方法，其特征在于，所述预设二次识别模型为预设逻辑回归模型，所述根据所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值，建立所述预设二次识别模型，包括：
    利用预设逻辑回归算法对所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值进行训练，建立所述预设逻辑回归模型。
  5. 根据权利要求3所述的方法，其特征在于，所述预设二次识别模型为预设神经网络模型，所述根据所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值，建立所述预设二次识别模型，包括：
    利用预设神经网络算法对所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值进行训练，建立所述预设神经网络模型。
  6. 根据权利要求3所述的方法，其特征在于，所述获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值，包括：
    将所述样本人群绩效特征输入到所述预设决策树模型进行识别，得到所述用户归属于对应分类人群绩效等级的概率值。
  7. 根据权利要求1所述的方法，其特征在于，所述将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值之前，所述方法还包括：
    获取待识别用户对应的人群绩效数据；
    从所述人群绩效数据中提取出所述用户对应的人群绩效特征。
  8. 一种人群绩效等级识别装置，其特征在于，包括：
    识别单元，用于将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
    所述识别单元，还用于将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
    确定单元，用于根据所述归属于各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
  9. 根据权利要求8所述的装置，其特征在于，所述确定单元，具体用于对所述各个人群绩效等级的概率值进行排序；将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。
  10. 根据权利要求8所述的装置，其特征在于，所述装置还包括：
    所述获取单元，用于获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级；
    所述建立单元，用于根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型；
    所述获取单元，还用于获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值；
    所述建立单元，还用于根据所述样本人群绩效特征和所述各个人样本群绩效等级的概率值，建立所述预设二次识别模型。
  11. 根据权利要求10所述的装置，其特征在于，所述建立单元，具体用于当所述预设二次识别模型为预设逻辑回归模型时，利用预设逻辑回归算法对所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值进行训练，建立所述预设逻辑 回归模型。
  12. 根据权利要求10所述的装置，其特征在于，所述建立单元，具体用于当所述预设二次识别模型为预设神经网络模型时，利用预设神经网络算法对所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值进行训练，建立所述预设神经网络模型。
  13. 根据权利要求10所述的装置，其特征在于，所述获取单元，具体用于将所述样本人群绩效特征输入到所述预设决策树模型进行识别，得到所述用户归属于对应分类人群绩效等级的概率值。
  14. 根据权利要求8所述的装置，其特征在于，所述装置还包括：
    获取单元，用于获取待识别用户对应的人群绩效数据；
    提取单元，用于从所述人群绩效数据中提取出所述用户对应的人群绩效特征。
  15. 一种计算机非易失性可读存储介质，其上存储有计算机可读指令，其特征在于，所述计算机可读指令被处理器执行时实现人群绩效等级识别方法，包括：
    将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
    将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
    根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
  16. 根据权利要求15所述的计算机非易失性可读存储介质，其特征在于，所述计算机可读指令被处理器执行时实现所述根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级，包括：
    对所述各个人群绩效等级的概率值进行排序；
    将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。
  17. 根据权利要求15所述的计算机非易失性可读存储介质，其特征在于，所述计算机可读指令被处理器执行时实现所述将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值之前，所述方法还包括：
    获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级；
    根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型，并获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值；
    根据所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值，建立所述预设二次识别模型。
  18. 一种计算机设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令，其特征在于，所述计算机可读指令被处理器执行时实现人群绩效等级识别方法，包括：
    将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值；
    将所述人群绩效特征和所述用户归属于对应分类人群绩效等级的概率值输入到预设二次识别模型进行二次识别，得到所述用户归属于各个人群绩效等级的概率值；
    根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级。
  19. 根据权利要求18所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时实现所述根据所述各个人群绩效等级的概率值，确定所述用户对应的人群绩效等级，包括：
    对所述各个人群绩效等级的概率值进行排序；
    将最高概率值对应的人群绩效等级，确定为所述用户对应的人群绩效等级。
  20. 根据权利要求18所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时实现所述将待识别用户对应的人群绩效特征输入到预设决策树模型进行一次识别，得到所述用户归属于对应分类人群绩效等级的概率值之前，所述方法还包括：
    获取多个样本用户对应的样本人群绩效特征和样本人群绩效等级；
    根据所述样本人群绩效特征和所述样本人群绩效等级，建立所述预设决策树模型，并获取所述预设决策树模型计算的所述多个样本用户归属于对应分类人群绩效等级的概率值；
    根据所述样本人群绩效特征和所述多个样本用户归属于对应分类人群绩效等级的概率值，建立所述预设二次识别模型。
PCT/CN2018/111124 2018-08-01 2018-10-21 人群绩效等级识别方法、装置、存储介质及计算机设备 Ceased WO2020024448A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810866432.9 2018-08-01
CN201810866432.9A CN109308565B (zh) 2018-08-01 2018-08-01 人群绩效等级识别方法、装置、存储介质及计算机设备

Publications (1)

Publication Number Publication Date
WO2020024448A1 true WO2020024448A1 (zh) 2020-02-06

Family

ID=65225980

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/111124 Ceased WO2020024448A1 (zh) 2018-08-01 2018-10-21 人群绩效等级识别方法、装置、存储介质及计算机设备

Country Status (2)

Country Link
CN (1) CN109308565B (zh)
WO (1) WO2020024448A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112817563B (zh) * 2020-03-26 2023-09-29 腾讯科技（深圳）有限公司 目标属性配置信息确定方法、计算机设备和存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104123395A (zh) * 2014-08-13 2014-10-29 北京赛科世纪数码科技有限公司 一种基于大数据的决策方法和系统
CN105225135A (zh) * 2015-10-31 2016-01-06 广州华多网络科技有限公司 潜力客户识别方法以及装置
CN108073883A (zh) * 2016-11-11 2018-05-25 深圳云天励飞技术有限公司 大规模人群属性识别方法及装置
CN108256691A (zh) * 2018-02-08 2018-07-06 成都智宝大数据科技有限公司 还款概率预测模型构建方法及装置

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7831463B2 (en) * 2004-09-07 2010-11-09 Sap Ag Computer-implemented method and system for allocating customer demand to suppliers
CN103412888B (zh) * 2013-07-19 2017-12-12 百度在线网络技术（北京）有限公司 一种兴趣点识别方法和装置
CN104778250B (zh) * 2015-04-14 2017-09-29 南京邮电大学 基于遗传规划决策树的信息物理融合系统数据分类方法
US20170083920A1 (en) * 2015-09-21 2017-03-23 Fair Isaac Corporation Hybrid method of decision tree and clustering technology
CN107665353A (zh) * 2017-09-15 2018-02-06 平安科技（深圳）有限公司 基于卷积神经网络的车型识别方法、装置、设备及计算机可读存储介质
CN107784482A (zh) * 2017-09-30 2018-03-09 平安科技（深圳）有限公司 招聘方法、电子装置及可读存储介质
CN108345880B (zh) * 2018-01-26 2021-07-27 金蝶软件（中国）有限公司 发票识别方法、装置、计算机设备和存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104123395A (zh) * 2014-08-13 2014-10-29 北京赛科世纪数码科技有限公司 一种基于大数据的决策方法和系统
CN105225135A (zh) * 2015-10-31 2016-01-06 广州华多网络科技有限公司 潜力客户识别方法以及装置
CN108073883A (zh) * 2016-11-11 2018-05-25 深圳云天励飞技术有限公司 大规模人群属性识别方法及装置
CN108256691A (zh) * 2018-02-08 2018-07-06 成都智宝大数据科技有限公司 还款概率预测模型构建方法及装置

Also Published As

Publication number Publication date
CN109308565B (zh) 2024-03-19
CN109308565A (zh) 2019-02-05

Similar Documents

Publication Publication Date Title
CN113326377B (zh) 一种基于企业关联关系的人名消歧方法及系统
WO2020024444A1 (zh) 人群绩效等级识别方法、装置、存储介质及计算机设备
WO2021155706A1 (zh) 利用不平衡正负样本对业务预测模型训练的方法及装置
CN107766929B (zh) 模型分析方法及装置
CN103914494B (zh) 一种微博用户身份识别方法及系统
WO2018014610A1 (zh) 基于c4.5决策树算法的特定用户挖掘系统及其方法
CN105373606A (zh) 一种改进c4.5决策树算法下的不平衡数据抽样方法
CN111754345A (zh) 一种基于改进随机森林的比特币地址分类方法
CN112418656A (zh) 智能坐席分派方法、装置、计算机设备和存储介质
CN108898476A (zh) 一种贷款客户信用评分方法和装置
WO2019061187A1 (zh) 信用评价方法和装置以及梯度渐进决策树参数调整方法和装置
CN110348516B (zh) 数据处理方法、装置、存储介质及电子设备
CN106408423A (zh) 用于风险评估的方法、系统及构建风险评估系统的方法
CN111159481B (zh) 图数据的边预测方法、装置及终端设备
CN110135681A (zh) 风险用户识别方法、装置、可读存储介质及终端设备
CN104679911B (zh) 一种基于离散弱相关的云平台决策森林分类方法
CN105306296A (zh) 一种基于lte信令的数据过滤处理方法
CN108629375A (zh) 电力客户分类方法、系统、终端及计算机可读存储介质
CN113742495B (zh) 基于预测模型的评级特征权重确定方法及装置、电子设备
CN109992592B (zh) 基于校园消费卡流水数据的高校贫困生识别方法
CN113723522B (zh) 异常用户的识别方法、装置、电子设备以及存储介质
CN115511073A (zh) 一种语义匹配模型的训练方法以及文本匹配方法
CN111491300A (zh) 风险检测方法、装置、设备及存储介质
CN115034509B (zh) 数据处理方法、装置、设备及存储介质
WO2020024448A1 (zh) 人群绩效等级识别方法、装置、存储介质及计算机设备

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: 18928669

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: 18928669

Country of ref document: EP

Kind code of ref document: A1