Customer group risk analysis method, device and system and medium

By determining and matching samples to eliminate the impact of confusing variables, the accuracy of customer risk analysis is improved, and the analysis inaccuracy problem caused by selective deviation in traditional methods is solved, and the accurate evaluation of the impact of intervention variables is achieved.

CN119048250APending Publication Date: 2024-11-29PING AN TECH (SHENZHEN) CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411235404.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-04
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Traditional customer risk analysis methods ignore selective bias factors, which leads to the difference in compensation risk between the intervention group and the control group that cannot be directly explained by the changes in characteristic indicators, reducing the accuracy and interpretability of the analysis.

Method used

By determining the intervention variables and confusion variables, dividing and matching samples, building new intervention groups and control groups, eliminating the impact of confusion variables, and conducting risk indicator differences analysis to obtain the results of the impact analysis of intervention variables on risk indicators.

Benefits of technology

The accuracy of customer group risk analysis is improved, and the interference of structural differences on risk analysis results can be eliminated, so as to more accurately evaluate the causal impact of intervention variables on compensation risk.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048250A_ABST
    Figure CN119048250A_ABST
Patent Text Reader

Abstract

The invention discloses a customer group risk analysis method, device and system and a medium, and the method comprises the steps: determining an intervention variable and a confusion variable according to a business scene, the intervention variable being used for representing a customer group feature difference, and the confusion variable being used for representing a customer group structure difference; dividing the customer group samples into an intervention group and a control group according to the intervention variables; performing tendency score prediction on the customer group samples according to the confusion variables, performing sample matching on the intervention group and the control group based on a tendency score prediction result, and constructing a new intervention group and a new control group; and carrying out risk index difference analysis on the new intervention group and the control group to obtain an influence analysis result of the intervention variables on the risk indexes. When the influence of the intervention variables on the risk indexes is analyzed, difference analysis is performed after the new intervention group and the new control group are constructed based on the tendency scores, so that the influence of confusion variables on risk differences can be eliminated, and the accuracy of customer group risk analysis is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, applicable to the financial field, and in particular to a customer risk analysis method, device, system and medium. Background Art

[0002] In the financial insurance industry, identifying customer risks is an important issue in insurance business operations. Identifying the claims risks of different customer groups helps insurance companies strengthen their differentiation in pricing, claims settlement and other aspects, and generate differentiated plans for different customer groups.

[0003] At present, the analysis of customer group claims risk often adopts the traditional method, which is to screen and segment the overall customers with certain characteristic indicators, obtain qualified intervention groups and control groups, and conduct claims risk analysis between different groups. However, the traditional method often ignores the impact of the factor of selective bias, so that the difference in claims risk between the intervention group and the control group cannot be directly explained by the change of characteristic indicators, which reduces the accuracy and interpretability of customer group risk analysis. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a customer risk analysis method, device, system and medium that can be applied to financial technology or other related fields, and its main purpose is to improve the accuracy of customer risk analysis.

[0005] The technical solution of the present invention is as follows:

[0006] A first aspect of the present invention provides a customer group risk analysis method, comprising:

[0007] Determine intervention variables and confounding variables based on business scenarios. The intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure.

[0008] Divide the sample of customers into an intervention group and a control group according to the intervention variables;

[0009] Predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing a new intervention group and a control group;

[0010] The difference analysis of risk indicators is performed on the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

[0011] In one embodiment, the predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing new intervention group and control group include:

[0012] The probability of each customer sample entering the intervention group is calculated based on the confounding variables to obtain the propensity score;

[0013] Based on the propensity score of each customer group sample, matching samples were screened in the intervention group and the control group according to the preset matching strategy to construct new intervention groups and control groups.

[0014] In one embodiment, before the method screens mutually matching samples in the intervention group and the control group based on the propensity score of each customer group sample according to a preset matching strategy and constructs new intervention group and control group, the method further comprises:

[0015] Sample pruning is performed based on the propensity score of each customer group sample to eliminate extreme samples.

[0016] In one embodiment, the preset matching strategy includes at least one of a nearest neighbor matching strategy, a radius matching strategy with boundary restrictions, and a hierarchical interval matching strategy.

[0017] In one embodiment, performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator includes:

[0018] Calculating first statistical data and second statistical data of the new intervention group and control group under the risk indicator respectively;

[0019] The first statistical data is compared with the second statistical data to obtain an analysis result of the impact of the intervention variable on the risk indicator.

[0020] In one embodiment, before performing differential analysis of risk indicators on the new intervention group and the control group, the method further comprises:

[0021] A matching balance test was performed on the new intervention group and the control group.

[0022] In one embodiment, after performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator, the method further comprises:

[0023] A sensitivity test was performed on the results of the analysis of the impact of the intervention variables on the risk indicators.

[0024] A second aspect of the present invention provides a customer group risk analysis device, comprising:

[0025] A variable determination module, used to determine intervention variables and confounding variables according to business scenarios, wherein the intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure;

[0026] A sample division module, used to divide the customer sample into an intervention group and a control group according to the intervention variable;

[0027] A propensity matching and construction module, used to predict the propensity score of the customer group sample according to the confounding variable, and to match the samples of the intervention group and the control group based on the propensity score prediction results to construct new intervention group and control group;

[0028] The risk analysis module is used to perform a difference analysis of the risk indicators of the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

[0029] A third aspect of the present invention provides a customer group risk analysis system, the system comprising at least one processor; and

[0030] a memory communicatively connected to the at least one processor; wherein,

[0031] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the above-mentioned customer group risk analysis method.

[0032] A fourth aspect of the present invention provides a non-volatile computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by one or more processors, the one or more processors can execute the above-mentioned customer risk analysis method.

[0033] Beneficial effects: The present invention discloses a method, device, system and medium for analyzing customer group risk. Compared with the prior art, the embodiment of the present invention determines intervention variables and confounding variables according to business scenarios, wherein the intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure; the customer group samples are divided into an intervention group and a control group according to the intervention variables; the customer group samples are predicted for propensity scores according to the confounding variables, and the intervention group and the control group are matched based on the prediction results of the propensity scores to construct new intervention groups and control groups; the new intervention group and the control group are analyzed for differences in risk indicators to obtain analysis results of the impact of the intervention variables on the risk indicators. By constructing new intervention groups and control groups based on propensity scores when analyzing the impact of intervention variables on risk indicators and then performing difference analysis, the impact of confounding variables on risk differences can be eliminated, thereby effectively improving the accuracy of customer group risk analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0035] Figure 1A flow chart of a customer group risk analysis method provided by an embodiment of the present invention;

[0036] Figure 2 A flow chart of step S103 in the customer group risk analysis method provided by an embodiment of the present invention;

[0037] Figure 3 A flow chart of step S104 in the customer group risk analysis method provided by an embodiment of the present invention;

[0038] Figure 4 A schematic diagram of the functional modules of a customer group risk analysis device provided by an embodiment of the present invention;

[0039] Figure 5 A schematic diagram of the hardware structure of the customer risk analysis system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail below. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. The embodiments of the present invention are described below in conjunction with the accompanying drawings.

[0041] In the financial insurance industry, identifying customer risks is an important issue in insurance business operations. Identifying the claims risks of different customer groups helps insurance companies strengthen their differentiation in pricing, claims settlement and other aspects, and generate differentiated plans for different customer groups.

[0042] At present, the analysis of customer group claims risk often adopts traditional methods, which uses certain characteristic indicators to screen and segment the overall customers, obtain qualified intervention groups and control groups, and conduct claims risk analysis between different groups. However, traditional methods often ignore the impact of the factor of selective bias. For example, the target customer group is customers with two cars, and the benchmark customer group is customers with one car. The two groups of customers may have structural differences in characteristics such as age and asset level, and structural factors will also affect the level of claims. Therefore, directly comparing the mean of claims cannot directly draw a conclusion that the characteristics of the two cars have a purely increased / reduced effect on the risk of the customer group, so that the difference in claims risk between the intervention group and the control group cannot be directly explained by the changes in characteristic indicators, which reduces the accuracy and interpretability of customer group risk analysis.

[0043] In order to solve the above problems, the present invention proposes a customer group risk analysis method, please refer to Figure 1 , Figure 1The present invention provides a flowchart of an embodiment of the customer group risk analysis method. The system architecture of the customer group risk analysis method provided in this embodiment may include a first terminal device, a second terminal device, a third terminal device, a network and a server. The network is used to provide a medium for a communication link between the first terminal device, the second terminal device, the third terminal device and the server. The network may include various connection types, such as wired and / or wireless communication links, etc.

[0044] The user may use the first terminal device, the second terminal device, and the third terminal device to interact with the server through the network to receive or send messages, etc. Various communication client applications may be installed on the first terminal device, the second terminal device, and the third terminal device, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only as examples).

[0045] The first terminal device, the second terminal device, and the third terminal device may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0046] The server may be a server that provides various services, such as a background management server that provides support for the content browsed by the user using the first terminal device, the second terminal device, and the third terminal device (for example only). The background management server may analyze and process the received data such as the user request, and feed back the processing results (such as web pages, information, or data obtained or generated according to the user request) to the terminal device. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server for a distributed system, or a server combined with a blockchain.

[0047] It should be noted that the customer group risk analysis method provided in the embodiment of the present invention can generally be executed by the first terminal device, the second terminal device or the third terminal device. Correspondingly, the customer group risk analysis device provided in the embodiment of the present invention can also be set in the first terminal device, the second terminal device or the third terminal device.

[0048] Alternatively, the customer group risk analysis method provided in the embodiment of the present invention may also be generally executed by a server. Accordingly, the customer group risk analysis device provided in the embodiment of the present invention may generally be arranged in a server. The customer group risk analysis method provided in the embodiment of the present invention may also be executed by a server or server cluster that is different from the server and can communicate with the first terminal device, the second terminal device, the third terminal device and / or the server. Accordingly, the customer group risk analysis device provided in the embodiment of the present invention may also be arranged in a server or server cluster that is different from the server and can communicate with the first terminal device, the second terminal device, the third terminal device and / or the server.

[0049] It should be understood that the numbers of the above terminal devices, networks and servers are only illustrative. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0050] like Figure 1 As shown, the customer group risk analysis method provided in this embodiment specifically includes the following steps:

[0051] S101. Determine intervention variables and confounding variables according to business scenarios, where the intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structures.

[0052] In this embodiment, when evaluating the impact of different variables on risk indicators, the corresponding intervention variables and confounding variables are determined according to the business scenario. The specific business scenario can be, for example, an auto insurance business scenario, a life insurance business scenario, a medical insurance business scenario, etc. The intervention variables used to characterize the differences in customer characteristics and the confounding variables used to characterize the differences in customer structure are determined according to the risk association characteristics of different business scenarios. For example, when evaluating the claim risk problem in the auto insurance business scenario, the claim ratio, accident rate, etc. are used as risk indicators (also called target variables) for comparing the claim risks of different customer groups. The intervention variables are the characteristic variables to be evaluated, that is, the factors whose impact on the target variables needs to be evaluated, such as whether the customer owns multiple cars; the confounding variables are the variables that affect the customer structure, such as gender, age, vehicle age, vehicle purchase price, etc., which can be selected based on business experience and scenarios, or automatically selected and processed after machine learning feature selection. When performing customer group risk analysis, this embodiment decomposes the influencing factors of risk differences into two parts, namely, differences caused by customer group structure and differences caused by customer group characteristics. Since the risk differences caused by customer group structure are actually a manifestation of selective bias, the impact of selective bias in customer group risk comparison can be eliminated through variable decomposition.

[0053] Preferably, the screened confounding variables can be further preprocessed, such as saturation cleaning and correlation screening, where saturation cleaning is to identify and remove variables with the same or almost the same values ​​in most samples, because these variables have very limited ability to distinguish different samples and therefore have limited effect on claims risk analysis; correlation screening is to identify confounding variables that are highly correlated with the target variable, for example, by calculating the correlation coefficient between each confounding variable and the target variable. Screening out highly correlated variables can help better analyze claims risks, and can further check the correlation between confounding variables to avoid multicollinearity problems caused by high correlations between multiple variables, thereby improving the accuracy of risk analysis.

[0054] S102. Divide the customer sample into an intervention group and a control group according to the intervention variable.

[0055] In this embodiment, after determining the intervention variables that need to analyze the difference in risk indicators, the pre-collected customer samples are divided according to the intervention variables to obtain the intervention group and the control group. Specifically, according to business needs, relevant customer data and insurance data, such as basic customer information, vehicle information, insurance purchase records, compensation history, accident records, etc., are collected. After the customer sample is constructed based on the collected data, the customer sample is divided into the intervention group and the control group based on the value of the intervention variable, where the intervention group refers to customers who meet the conditions of the intervention variables, such as customers who own multiple cars, and the control group refers to customers who do not meet the conditions of the intervention variables, such as customers who only own one car. The intervention group and the control group are divided by the intervention variables to evaluate the impact of the intervention variables on risk indicators such as the claim rate and the accident rate.

[0056] S103, predicting the propensity score of the customer group sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing a new intervention group and a control group.

[0057] In this embodiment, in order to eliminate the influence of selective bias in the customer group risk comparison problem, after the initial intervention group and control group are divided, the customer group samples are also predicted according to the confounding variables. The propensity score prediction represents the conditional probability of each customer group sample accepting intervention (for example, purchasing a certain auto insurance product or enjoying a certain preferential policy) according to its confounding variables (such as age, gender, vehicle age, etc.). Therefore, by predicting the propensity score of each customer group sample, the influence of the confounding variable on the risk index can be controlled and the selection bias can be reduced. When reducing the selection bias, the propensity score prediction results are specifically used to match the samples in the initial divided intervention group and control group to construct a new intervention group and control group. The influence of the confounding variables on the results of the two groups of samples is consistent. As a result, the difference between the newly constructed intervention group and control group is not affected by the confounding variables, and the actual influence of the intervention variables on the risk indicators can be compared during risk analysis.

[0058] In one embodiment, Figure 2 As shown, the propensity score of the customer group sample is predicted according to the confounding variable, and the intervention group and the control group are matched based on the propensity score prediction result to construct a new intervention group and a control group, including:

[0059] S201, calculating the probability of each customer group sample entering the intervention group according to the confounding variable to obtain a propensity score;

[0060] S202. Based on the propensity score of each customer group sample, mutually matching samples are screened in the intervention group and the control group according to a preset matching strategy to construct new intervention group and control group.

[0061] In this embodiment, when constructing new intervention and control groups, the probability of each customer sample entering the intervention group is first calculated based on the confounding variables to obtain the propensity score. Specifically, a machine learning model can be established and trained to calculate the propensity score, such as linear regression, logistic regression, LightGBM and other models, and then the intervention variable is used as the dependent variable in the model, that is, the predicted variable, and the confounding variable is used as the independent variable of the model, that is, the variable that may affect the intervention allocation. The model is trained by collecting and preprocessing the obtained training samples, and the conditional probability of each sample receiving the intervention is estimated. After the training is completed, the trained model can be used to calculate the probability of each sample becoming an intervention group based on the currently screened confounding variables. This probability value is the propensity score of the sample, and the calculation result of the propensity score can be used for subsequent matching analysis to evaluate the causal effect of the intervention.

[0062] Based on the propensity score of each customer group sample, samples with matching scores are selected in the intervention group and the control group according to the preset matching strategy to establish new intervention groups and control groups, so that the influence of confounding variables on the results of the new two groups of samples is consistent, thereby eliminating the influence of selective bias in risk analysis. The specific matching strategy is at least one of the nearest neighbor matching strategy, the radius matching strategy with boundary restrictions, and the stratified interval matching strategy, where the nearest neighbor matching refers to selecting users in the intervention group with the smallest difference in propensity scores in the control group for matching; the radius matching with boundary restrictions refers to tolerating the difference in propensity scores between the intervention group and the control group within a certain radius; the stratified interval matching refers to dividing the propensity score into multiple intervals and matching in each interval. The specific matching method can be flexibly selected and set according to the needs, and this embodiment does not limit this.

[0063] In one embodiment, before the method screens mutually matching samples in the intervention group and the control group based on the propensity score of each customer group sample according to a preset matching strategy and constructs new intervention group and control group, the method further comprises:

[0064] Sample pruning is performed based on the propensity score of each customer group sample to eliminate extreme samples.

[0065] In this embodiment, before matching and establishing a new intervention group and control group, sample pruning is also performed based on the propensity score of each customer group sample to improve the matching quality and the accuracy of the estimation. Specifically, a suitable pruning threshold can be determined based on the distribution of the propensity score, for example, based on the highest score and the lowest score, the score of a specific percentage (such as 1% or 5%) from it is used as the pruning threshold to delete samples with extreme propensity score distribution at both ends; or the intersection of the propensity scores of the intervention group and the control group can be obtained, and the score corresponding to the specified proportion (such as 90%-95%) in the middle of the interval of the intersection is used as the pruning threshold to delete samples with propensity scores outside the vast majority of the intervals of the intersection of the intervention group and the control group. Using the pruned samples for subsequent matching analysis can reduce the influence of the extreme values ​​at both ends of the propensity score distribution on the matching results, so that the pruned samples can more easily find matching pairs and improve the matching quality.

[0066] S104, performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator.

[0067] In this embodiment, based on the intervention group and control group reconstructed after matching, the difference analysis of risk indicators is performed. Since the influence of the confounding variables on the corresponding analysis indicators of the new intervention group and the control group is consistent, the difference analysis of risk indicators of the two sample groups at this time will not be affected by the confounding variables. By comparing the differences in risk indicators such as the claim ratio and the accident rate between the new intervention group and the control group, the analysis results of the influence of the intervention variables on the risk indicators can be obtained, such as the increase or decrease of the claim ratio after the intervention of the customer group characteristics by the intervention variables, the increase or decrease, and the magnitude of the increase or decrease. By combining the propensity score matching to construct a new intervention group and control group, the interference of the confounding variables, i.e., the differences in the customer group structure, on the risk analysis results is removed during the customer group claim risk analysis, and the intervention group and the control group are comparable in the dimension of customer group attributes, so that the actual effect of the intervention variables on the claim risk can be directly evaluated, and the causal influence of the intervention variables on the claim risk indicators can be more accurately inferred, thereby improving the accuracy of risk analysis, helping to identify customer group risks in order to provide targeted services and solutions.

[0068] In one embodiment, Figure 3 As shown, the difference analysis of risk indicators is performed on the new intervention group and the control group to obtain the analysis results of the impact of the intervention variable on the risk indicator, including:

[0069] Calculating first statistical data and second statistical data of the new intervention group and control group under the risk indicator respectively;

[0070] The first statistical data is compared with the second statistical data to obtain an analysis result of the impact of the intervention variable on the risk indicator.

[0071] In this embodiment, when analyzing the risk differences between different customer groups, the first statistical data and the second statistical data of the new intervention group and the control group under the risk indicator are calculated respectively. For example, when the loss ratio is used as the risk indicator, an appropriate descriptive statistic, such as the mean, median, mode, etc., can be determined as a basis for statistical analysis; then, the loss ratio data of each customer group sample in the new intervention group and the control group are collected, and the loss ratio data can be further processed, such as cleaning the data, processing missing values ​​and outliers, etc., to improve statistical reliability; then, the loss ratio data of the new intervention group is statistically calculated to obtain the first statistical data, such as calculating the mean loss ratio of the new intervention group, and the loss ratio data of the new control group is statistically calculated to obtain the second statistical data, and the mean loss ratio of the new control group is also calculated.

[0072] Based on the calculated statistical data of the two groups of samples, the two groups of statistical data are compared to obtain the actual impact of the intervention variable on the risk index after eliminating the selective bias. Specifically, there are many ways of comparative analysis, such as direct difference comparison, by directly calculating the statistical data difference of the risk index of the intervention group and the control group, such as the mean difference or median difference, etc., to obtain the risk analysis result that the intervention variable has an increase or decrease in the risk index; or use statistical tests (such as t-test, Z-test, non-parametric test, etc.) to determine whether the difference between the two groups is statistically significant, so as to determine whether the intervention may have affected the risk index; or use box plots, histograms and other visualization tools to analyze the distribution of the two groups of data, compare the central tendency and dispersion of the two groups of statistical data, etc., and flexibly select the corresponding comparison method according to the actual analysis needs.

[0073] By analyzing the differences between the statistical data of the new intervention group and the control group, if the risk indicator statistics of the new intervention group are significantly different from those of the control group in both statistical and practical terms, it indicates that the intervention variable has an impact on the risk indicator. For example, if the mean claim ratio of the new intervention group is significantly lower than that of the control group, it may indicate that the intervention measures have effectively reduced the risk of claims; conversely, if the claim ratio of the new intervention group is higher than that of the control group, it may indicate that the intervention measures have no effect, or may even increase the risk. By attributing causal effects to customer risk analysis on intervention behaviors, the actual impact of intervention variables on risk indicators can be more accurately evaluated, effectively improving the reliability and accuracy of risk analysis.

[0074] In one embodiment, before performing differential analysis of risk indicators on the new intervention group and the control group, the method further comprises:

[0075] A matching balance test was performed on the new intervention group and the control group.

[0076] In this embodiment, before the new intervention group and control group are subjected to differential analysis of risk indicators, in order to ensure the reliability of the analysis results, the new intervention group and control group are first subjected to a matching balance test to ensure that the matched intervention group and control group are similarly distributed on the confounding variables, thereby reducing selection bias and estimation errors. Specifically, a matching balance test can be performed by methods such as standardized deviation, T test, and joint significance. For example, the standardized deviation is to calculate the mean difference between the new intervention group and control group on each confounding variable and standardize it. The absolute value of the standardized deviation (Standardized Difference) is less than 0.1 or 0.25 and is generally considered to be an acceptable balance level; the T test is to perform a T test or a Z test on each confounding variable to determine whether there is a statistically significant difference between the intervention group and the control group; the joint significance is to use a multiple regression model, use the propensity score as a covariate, and test the joint significance of multiple confounding variables at the same time. It can be flexibly selected according to demand, and this embodiment is not limited to this. The matching balance test ensures that the intervention group and control group constructed after matching are comparable on the confounding variables, so that the causal effect of risk analysis can be more reliably attributed to the intervention variable itself, rather than other confounding factors. If the matching balance test fails, that is, there are still significant differences between the two groups in the confounding variables, it is necessary to re-update the constructed intervention group and control group, for example, by re-matching the intervention group and control group through other matching methods to ensure the reliability of the subsequent risk difference analysis.

[0077] In one embodiment, after performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator, the method further comprises:

[0078] A sensitivity test was performed on the results of the analysis of the impact of the intervention variables on the risk indicators.

[0079] In this embodiment, in order to ensure the reliability and stability of the current analysis results, after performing a differential analysis of the risk indicators to obtain the analysis results of the impact of the intervention variables on the risk indicators, a sensitivity test is also performed on the impact analysis results. For example, an independent random variable can be added to see whether there is a significant impact on the result, or whether the causal effect is zeroed after replacing the intervention variable / target variable with a confounding variable, etc. The sensitivity test can be used to evaluate whether the differences in risk indicators (such as mean compensation) between the new intervention group and the control group are robust, that is, to ensure that the risk analysis results are not accidental due to specific model selection or assumptions, but remain robust within a certain range of changes, thereby ensuring the credibility of the risk analysis results.

[0080] It should be noted that there is not necessarily a certain order between the above-mentioned steps. A person skilled in the art can understand, based on the description of the embodiments of the present invention, that in different embodiments, the above-mentioned steps may have different execution orders, that is, they may be executed in parallel, may be executed interchangeably, and so on.

[0081] Another embodiment of the present invention provides a customer group risk analysis device, such as Figure 4 As shown, the device 1 comprises:

[0082] A variable determination module 11 is used to determine intervention variables and confounding variables according to business scenarios, wherein the intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure;

[0083] A sample division module 12 is used to divide the customer sample into an intervention group and a control group according to the intervention variable;

[0084] The propensity matching and construction module 13 is used to predict the propensity score of the customer group sample according to the confounding variable, and to match the samples of the intervention group and the control group based on the propensity score prediction results to construct new intervention group and control group;

[0085] The risk analysis module 14 is used to perform a difference analysis of the risk indicators of the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

[0086] The module referred to in the present invention refers to a series of computer program instruction segments that can complete specific functions, which are more suitable for describing the execution process of customer risk analysis than programs. For the specific implementation methods of each module, please refer to the corresponding method embodiments above, which will not be repeated here.

[0087] In one embodiment, the propensity matching and construction module 13 includes:

[0088] A propensity score calculation unit, used to calculate the probability of each customer group sample entering the intervention group according to the confounding variable to obtain a propensity score;

[0089] The matching construction unit is used to screen mutually matching samples in the intervention group and the control group according to the preset matching strategy based on the propensity score of each customer group sample, and construct new intervention group and control group.

[0090] In one embodiment, the device 1 further comprises:

[0091] The sample pruning module is used to prune samples according to the propensity score of each customer group sample and eliminate extreme samples.

[0092] In one embodiment, the preset matching strategy includes at least one of a nearest neighbor matching strategy, a radius matching strategy with boundary restrictions, and a hierarchical interval matching strategy.

[0093] In one embodiment, the risk analysis module 14 includes:

[0094] A risk indicator statistical unit, used to calculate first statistical data and second statistical data of the new intervention group and control group under the risk indicator respectively;

[0095] The comparison and analysis unit is used to compare the first statistical data with the second statistical data to obtain an analysis result of the impact of the intervention variable on the risk indicator.

[0096] In one embodiment, the device 1 further comprises:

[0097] The first test module is used to perform a matching balance test on the new intervention group and the control group.

[0098] In one embodiment, the device 1 further comprises:

[0099] The second test module is used to perform a sensitivity test on the analysis result of the impact of the intervention variable on the risk indicator.

[0100] Another embodiment of the present invention provides a customer group risk analysis system, such as Figure 5 As shown, the system 10 includes:

[0101] One or more processors 110 and memory 120, Figure 5 A processor 110 is used as an example for description. The processor 110 and the memory 120 may be connected via a bus or other means. Figure 5 The example of connecting through bus is taken in the following.

[0102] The processor 110 is used to complete various control logics of the system 10, and it can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a single-chip microcomputer, an ARM (Acorn RISC Machine) or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or any combination of these components. In addition, the processor 110 can also be any traditional processor, microprocessor or state machine. The processor 110 can also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a DSP and / or any other such configuration.

[0103] The memory 120 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions corresponding to the customer risk analysis method in the embodiment of the present invention. The processor 110 executes various functional applications and data processing of the system 10 by running the non-volatile software programs, instructions and units stored in the memory 120, that is, implements the customer risk analysis method in the above method embodiment.

[0104] The memory 120 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required by at least one function; the data storage area may store data created according to the use of the system 10, etc. In addition, the memory 120 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 120 may optionally include a memory remotely arranged relative to the processor 110, and these remote memories may be connected to the system 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0105] One or more units are stored in the memory 120, and when executed by one or more processors 110, the customer group risk analysis method in any of the above method embodiments is executed, for example, the following steps are executed:

[0106] Determine intervention variables and confounding variables based on business scenarios. The intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure.

[0107] Divide the sample of customers into an intervention group and a control group according to the intervention variables;

[0108] Predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing a new intervention group and a control group;

[0109] The difference analysis of risk indicators is performed on the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

[0110] In one embodiment, the predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing new intervention group and control group include:

[0111] The probability of each customer sample entering the intervention group is calculated based on the confounding variables to obtain the propensity score;

[0112] Based on the propensity score of each customer group sample, matching samples were screened in the intervention group and the control group according to the preset matching strategy to construct new intervention groups and control groups.

[0113] In one embodiment, before the method screens mutually matching samples in the intervention group and the control group based on the propensity score of each customer group sample according to a preset matching strategy and constructs new intervention group and control group, the method further comprises:

[0114] Sample pruning is performed based on the propensity score of each customer group sample to eliminate extreme samples.

[0115] In one embodiment, the preset matching strategy includes at least one of a nearest neighbor matching strategy, a radius matching strategy with boundary restrictions, and a hierarchical interval matching strategy.

[0116] In one embodiment, performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator includes:

[0117] Calculating first statistical data and second statistical data of the new intervention group and control group under the risk indicator respectively;

[0118] The first statistical data is compared with the second statistical data to obtain an analysis result of the impact of the intervention variable on the risk indicator.

[0119] In one embodiment, before performing differential analysis of risk indicators on the new intervention group and the control group, the method further comprises:

[0120] A matching balance test was performed on the new intervention group and the control group.

[0121] In one embodiment, after performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator, the method further comprises:

[0122] A sensitivity test was performed on the results of the analysis of the impact of the intervention variables on the risk indicators.

[0123] An embodiment of the present invention provides a non-volatile computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by one or more processors, the customer group risk analysis method in any of the above method embodiments is executed, for example, the following steps are performed:

[0124] Determine intervention variables and confounding variables based on business scenarios. The intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure.

[0125] Divide the sample of customers into an intervention group and a control group according to the intervention variables;

[0126] Predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing a new intervention group and a control group;

[0127] The difference analysis of risk indicators is performed on the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

[0128] In one embodiment, the predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing new intervention group and control group include:

[0129] The probability of each customer sample entering the intervention group is calculated based on the confounding variables to obtain the propensity score;

[0130] Based on the propensity score of each customer group sample, matching samples were screened in the intervention group and the control group according to the preset matching strategy to construct new intervention groups and control groups.

[0131] In one embodiment, before the method screens mutually matching samples in the intervention group and the control group based on the propensity score of each customer group sample according to a preset matching strategy and constructs new intervention group and control group, the method further comprises:

[0132] Sample pruning is performed based on the propensity score of each customer group sample to eliminate extreme samples.

[0133] In one embodiment, the preset matching strategy includes at least one of a nearest neighbor matching strategy, a radius matching strategy with boundary restrictions, and a hierarchical interval matching strategy.

[0134] In one embodiment, performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator includes:

[0135] Calculating first statistical data and second statistical data of the new intervention group and control group under the risk indicator respectively;

[0136] The first statistical data is compared with the second statistical data to obtain an analysis result of the impact of the intervention variable on the risk indicator.

[0137] In one embodiment, before performing differential analysis of risk indicators on the new intervention group and the control group, the method further comprises:

[0138] A matching balance test was performed on the new intervention group and the control group.

[0139] In one embodiment, after performing a difference analysis of risk indicators on the new intervention group and the control group to obtain an analysis result of the impact of the intervention variable on the risk indicator, the method further comprises:

[0140] A sensitivity test was performed on the results of the analysis of the impact of the intervention variables on the risk indicators.

[0141] As an example, non-volatile storage media can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) as external cache memory. By way of illustration and not limitation, RAM can be obtained in many forms such as synchronous RAM (SRAM), dynamic RAM, (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The disclosed memory components or memories of the operating environment described herein are intended to include one or more of these and / or any other suitable types of memory.

[0142] In summary, the present invention discloses a method, device, system and medium for analyzing customer group risk, wherein the method includes: determining intervention variables and confounding variables according to business scenarios, wherein the intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure; dividing customer group samples into an intervention group and a control group according to the intervention variables; predicting the propensity score of the customer group samples according to the confounding variables, performing sample matching on the intervention group and the control group based on the propensity score prediction results, and constructing new intervention groups and control groups; performing differential analysis on risk indicators of the new intervention group and control group, and obtaining analysis results of the impact of the intervention variables on the risk indicators. By constructing new intervention groups and control groups based on propensity scores and performing differential analysis when analyzing the impact of intervention variables on risk indicators, the impact of confounding variables on risk differences can be eliminated, thereby effectively improving the accuracy of customer group risk analysis.

[0143] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium, and the computer program can include the processes of the above-mentioned method embodiments when executed. The storage medium can be a memory, a disk, a floppy disk, a flash memory, an optical storage device, etc.

[0144] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A customer group risk analysis method, characterized in that: include: Determine intervention variables and confounding variables based on business scenarios. The intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure. Divide the sample of customers into an intervention group and a control group according to the intervention variables; Predicting the propensity score of the customer sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing a new intervention group and a control group; The difference analysis of risk indicators is performed on the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

2. The customer group risk analysis method according to claim 1, characterized in that: The method of predicting the propensity score of the customer group sample according to the confounding variable, performing sample matching on the intervention group and the control group based on the propensity score prediction result, and constructing a new intervention group and a control group includes: The probability of each customer sample entering the intervention group is calculated based on the confounding variables to obtain the propensity score; Based on the propensity score of each customer group sample, matching samples were screened in the intervention group and the control group according to the preset matching strategy to construct new intervention groups and control groups.

3. The customer group risk analysis method according to claim 2, characterized in that: Before the method screens mutually matching samples in the intervention group and the control group based on the propensity score of each customer group sample according to a preset matching strategy to construct new intervention groups and control groups, the method further comprises: Sample pruning is performed based on the propensity score of each customer group sample to eliminate extreme samples.

4. The customer group risk analysis method according to claim 2, characterized in that: The preset matching strategy includes at least one of a nearest neighbor matching strategy, a radius matching strategy with boundary restrictions, and a hierarchical interval matching strategy.

5. The customer group risk analysis method according to claim 1, characterized in that: The difference analysis of the risk indicators of the new intervention group and the control group is performed to obtain the analysis results of the impact of the intervention variables on the risk indicators, including: Calculating first statistical data and second statistical data of the new intervention group and control group under the risk indicator respectively; The first statistical data is compared with the second statistical data to obtain an analysis result of the impact of the intervention variable on the risk indicator.

6. The customer group risk analysis method according to claim 1, characterized in that: Before performing differential analysis of risk indicators on the new intervention group and the control group, the method further comprises: A matching balance test was performed on the new intervention group and the control group.

7. The customer group risk analysis method according to claim 1, characterized in that: After performing a difference analysis of the risk indicators on the new intervention group and the control group to obtain the analysis results of the impact of the intervention variable on the risk indicator, the method further includes: A sensitivity test was performed on the results of the analysis of the impact of the intervention variables on the risk indicators.

8. A customer group risk analysis device, characterized in that: include: A variable determination module, used to determine intervention variables and confounding variables according to business scenarios, wherein the intervention variables are used to characterize differences in customer group characteristics, and the confounding variables are used to characterize differences in customer group structure; A sample division module, used to divide the customer sample into an intervention group and a control group according to the intervention variable; A propensity matching and construction module, used to predict the propensity score of the customer group sample according to the confounding variable, and to match the samples of the intervention group and the control group based on the propensity score prediction results to construct new intervention group and control group; The risk analysis module is used to perform a difference analysis of the risk indicators of the new intervention group and the control group to obtain the analysis results of the impact of the intervention variables on the risk indicators.

9. A customer group risk analysis system, characterized in that: The system includes at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the customer group risk analysis method according to any one of claims 1 to 7.

10. A non-volatile computer-readable storage medium, characterized in that: The non-volatile computer-readable storage medium stores computer-executable instructions, which, when executed by one or more processors, enable the one or more processors to execute the customer risk analysis method according to any one of claims 1 to 7.

Citation Information

Cited By

  • Method for recognizing VOCs biomarkers in expired air of physiologically abnormal crowd based on machine learning and application

    CN119724371A