Risk list generation method and device, equipment, medium and product
By using the random forest Bagging algorithm to generate a risk prediction model in the non-vehicle claims data processing, the problem of difficulty in efficiently identifying non-vehicle claims data risks in the prior art is solved, and efficient and accurate risk management is achieved.
Patent Information
- Application Number
- CN202411670493.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-20
- Publication Date
- 2025-02-28
AI Technical Summary
It is difficult for the prior art to efficiently determine the data risk situation in non-vehicle claims data. As the types and quantity of claims data expand, it is difficult to identify data risks in a timely manner.
The random forest Bagging algorithm is used to generate a risk prediction model based on the preset data set, and the corresponding risk prediction model is generated through feature extraction and multi-decision tree construction, and the target claim data is input to the model to generate a risk label, and finally a risk list is generated based on the label.
It realizes efficient identification of data risks and generates an accurate risk list when processing non-vehicle claims data, and improves the efficiency and accuracy of risk management.
Smart Images
Figure CN119539969A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of financial technology, and in particular to a method, device, equipment, medium and product for generating a risk list. Background Art
[0002] In the financial field, such as the insurance industry, the risk control list and risk control rules currently used in non-auto insurance claims are all from third parties. The risk control list comes from the blacklist of the underwriting side, and the risk control rules come from the qualitative rules formulated by the claims business.
[0003] Since the risk control list and risk control rules are provided by a third party on a fixed basis, as the types and quantities of claims data continue to expand, it is difficult to determine the data risk situation in non-auto claims data in a timely manner.
[0004] Therefore, a method is needed to efficiently determine the data risk situation in non-auto claims data. Summary of the invention
[0005] The present application provides a risk list generation method, device, equipment, medium and product to solve the current problem of needing a method for efficiently determining data risk conditions in non-auto claims data.
[0006] The first aspect of the present application provides a method for generating a risk list, comprising:
[0007] A random forest bagging algorithm is used to generate a corresponding risk prediction model based on a preset data set; the preset data set includes historical claims-related data; the historical claims-related data includes: customer information, claims information and policy information;
[0008] Obtain target claims data to be processed;
[0009] Inputting the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data;
[0010] A risk list corresponding to the target claims data is generated according to the risk tags.
[0011] Furthermore, in the above method, the random forest bagging algorithm is used to generate a corresponding risk prediction model based on a preset data set, including:
[0012] Perform feature extraction on the preset data set to generate a corresponding feature set;
[0013] A random forest bagging algorithm is used to construct a corresponding plurality of decision trees based on the feature set;
[0014] The corresponding risk prediction model is constructed based on the multiple decision trees finally constructed.
[0015] Furthermore, in the above method, the corresponding risk prediction model is constructed based on the multiple decision trees finally constructed, including:
[0016] A preset rule engine is used to configure the mapping relationship between the risk dimension prediction results generated by multiple decision trees and the risk labels to generate a corresponding risk prediction model; each risk dimension in the risk dimension prediction results is a pre-set dimensional data.
[0017] Furthermore, in the above method, the step of obtaining the target claims data to be processed includes:
[0018] Sending a claim data acquisition request to a claim-related electronic device;
[0019] Receiving corresponding initial claim data sent by electronic equipment related to claims;
[0020] The initial claims data is cleaned and preprocessed to generate target claims data.
[0021] Furthermore, in the above method, generating a risk list corresponding to the target claims data according to the risk tag includes:
[0022] Determining associated data corresponding to the risk tag in the target claims data;
[0023] Import the associated data into the preset list template to generate the corresponding risk list.
[0024] Furthermore, the method as described above further comprises:
[0025] Obtaining misjudgment list information from a preset database; the misjudgment list information is claim data that is not on the risk list and has risks or claim data that is on the risk list but has no risks;
[0026] The risk prediction model is retrained based on the misjudgment list information to generate an optimized risk prediction model.
[0027] Furthermore, the method as described above, after using a preset rule engine to configure a mapping relationship between risk dimension prediction results generated by multiple decision trees and risk labels to generate a corresponding risk prediction model, further includes:
[0028] Obtaining a test sample; the test sample includes test claims data;
[0029] Inputting the test sample into a risk prediction model to generate a test risk label corresponding to the test sample;
[0030] Compare the test risk label with the actual risk label; the actual risk label is the risk label manually annotated on the test claim data;
[0031] If the test risk label is the same as the actual risk label, the test result of the risk prediction model is determined to be a passed test.
[0032] A second aspect of the present application provides a risk list generating device, comprising:
[0033] The first generation module is used to generate a corresponding risk prediction model based on a preset data set using a random forest bagging algorithm; the preset data set includes historical claims-related data; the historical claims-related data includes: customer information, claims information and policy information;
[0034] An acquisition module is used to acquire target claims data to be processed;
[0035] An input module, used for inputting the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data;
[0036] The second generating module is used to generate a risk list corresponding to the target claim data according to the risk label.
[0037] Further, in the above device, the first generating module is specifically used for:
[0038] Perform feature extraction on the preset data set to generate a corresponding feature set; use the random forest bagging algorithm to construct corresponding multiple decision trees based on the feature set; and construct a corresponding risk prediction model based on the multiple decision trees finally constructed.
[0039] Furthermore, in the above-mentioned device, when the first generating module constructs the corresponding risk prediction model based on the multiple decision trees finally constructed, it is specifically used to:
[0040] A preset rule engine is used to configure the mapping relationship between the risk dimension prediction results generated by multiple decision trees and the risk labels to generate a corresponding risk prediction model; each risk dimension in the risk dimension prediction results is a pre-set dimensional data.
[0041] Furthermore, in the above device, the acquisition module is specifically used for:
[0042] Sending a claim data acquisition request to an electronic device related to claim settlement; receiving corresponding initial claim data sent by the electronic device related to claim settlement; performing data cleaning and data preprocessing on the initial claim data to generate target claim data.
[0043] Further, in the above device, the second generating module is specifically used for:
[0044] Determine the associated data corresponding to the risk tag in the target claims data; import the associated data into a preset list template to generate a corresponding risk list.
[0045] Furthermore, the device as described above further comprises:
[0046] A training module is used to obtain misjudgment list information from a preset database; the misjudgment list information is claim data that is not on the risk list and has risks, or claim data that is on the risk list but does not have risks; based on the misjudgment list information, the risk prediction model is re-trained to generate an optimized risk prediction model.
[0047] Furthermore, in the above device, the first generating module is further used for:
[0048] Obtain a test sample; the test sample includes test claims data; input the test sample into a risk prediction model to generate a test risk label corresponding to the test sample; compare the test risk label with the actual risk label; the actual risk label is a risk label manually annotated on the test claims data; if the test risk label is the same as the actual risk label, determine that the test result of the risk prediction model is a passed test.
[0049] A third aspect of the present application provides an electronic device, including: a memory and a processor;
[0050] The memory stores computer-executable instructions;
[0051] The processor executes the computer-executable instructions stored in the memory to implement the risk list generation method as described in any one of the first aspects.
[0052] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the risk list generation method described in any one of the first aspects.
[0053] A fifth aspect of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the risk list generation method described in any one of the first aspects.
[0054] The present application provides a method, device, equipment, medium and product for generating a risk list, the method comprising: using a random forest bagging algorithm to generate a corresponding risk prediction model based on a preset data set; the preset data set includes historical claims-related data; the historical claims-related data includes: customer information, claims information and policy information; obtaining the target claims data to be processed; inputting the target claims data into the risk prediction model to generate a risk label corresponding to the target claims data; generating a risk list corresponding to the target claims data according to the risk label. The risk list generation method of the present application generates a corresponding risk prediction model based on a preset data set through a random forest bagging algorithm, so that when processing the target claims data, a risk label corresponding to the target claims data can be generated based on the risk prediction model. At the same time, a risk list corresponding to the target claims data is generated according to the risk label, so that the data risk situation and the specific risk list in the claims data can be determined proactively and more efficiently. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0056] Figure 1 A scenario diagram for implementing the risk list generation method of the embodiment of the present application;
[0057] Figure 2 A flowchart of a method for generating a risk list provided in an embodiment of the present application;
[0058] Figure 3 A flowchart of a method for generating a risk list provided in another embodiment of the present application;
[0059] Figure 4 A schematic diagram of the structure of the risk list generation device provided in this application;
[0060] Figure 5 A schematic diagram of the structure of the electronic device provided in this application.
[0061] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0062] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.
[0063] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of financial data or user data and other information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals. It should be noted that in the embodiments of this application, some existing solutions in the industry such as certain software, components, models, etc. may be mentioned, and they should be considered as exemplary. Their purpose is only to illustrate the feasibility of the implementation of the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.
[0064] It should be noted that the disclosed risk list generation method, device, equipment, medium and product can be used in the field of financial technology. It can also be used in any field other than the field of financial technology. The application field of the disclosed risk list generation method, device, equipment, medium and product is not limited.
[0065] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.
[0066] In order to clearly understand the technical solution of this application, the invention concept of this application is first introduced in detail. At present, in the financial field, such as the insurance industry, the existing insurance risk list generation and management is mainly generated by analyzing underwriting data to generate black and gray lists, or the claims business department forms risk management conditions through post-qualitative analysis of claims data.
[0067] In order to achieve the purpose of reducing claims and losses and risk control, it is necessary to develop risk list generation technology for the claims system from the perspective of insurance affairs; from a technical perspective, the claims system has full and accurate claims production data and can better grasp the risk control points of claims. At the same time, since the risk control list and risk control rules are fixedly provided by a third party, as the types and quantities of claims data continue to expand, it is difficult to determine the data risk situation in non-auto claims data in a timely manner.
[0068] Therefore, a method is needed to efficiently determine the data risk situation in non-auto claims data.
[0069] Therefore, in view of the problem that the prior art requires a method to efficiently determine the data risk situation in non-auto claims data, the inventors found in their research that in order to solve this problem, a corresponding risk prediction model can be generated based on a preset data set through a random forest bagging algorithm, so that when processing target claims data, a risk label corresponding to the target claims data can be generated based on the risk prediction model. At the same time, a risk list corresponding to the target claims data is generated based on the risk label, so that the data risk situation and specific risk list in the claims data can be determined proactively and more efficiently.
[0070] Specifically, a random forest bagging algorithm is used to generate a corresponding risk prediction model based on a preset data set; the preset data set includes historical claims related data; the historical claims related data includes: customer information, claims information and policy information; the target claims data to be processed is obtained; the target claims data is input into the risk prediction model to generate a risk label corresponding to the target claims data; and a risk list corresponding to the target claims data is generated according to the risk label.
[0071] Based on the above creative findings, the inventor proposed the technical solution of the present application.
[0072] The following is an introduction to the application scenarios of the risk list generation method provided in the embodiment of the present application. Figure 1 As shown, 1 is a first electronic device, and 2 is a second electronic device. The first electronic device 1 can be a user terminal, an electronic device related to a claims settlement system, etc. The second electronic device 2 can be an Internet platform device, a database device, etc.
[0073] Exemplarily, when the first electronic device 1 monitors and analyzes the claims data to generate a risk list, the specific steps are as follows:
[0074] ① Use the random forest bagging algorithm to generate the corresponding risk prediction model based on the preset data set.
[0075] ② The second electronic device 2 sends the target claim data to the first electronic device 1.
[0076] The first electronic device 1 performs the following process:
[0077] ③ Input the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data.
[0078] ④ Generate a risk list corresponding to the target claims data based on the risk tags.
[0079] Optionally, the risk list may also be displayed to the user so that the user is aware of the risk list and can manually review the accuracy of the risk list again, and prepare for subsequent tuning of the risk prediction model.
[0080] The embodiments of the present application are introduced below in conjunction with the drawings in the specification.
[0081] Figure 2 A flow chart of a method for generating a risk list provided in an embodiment of the present application is shown as follows: Figure 2 As shown, the execution subject in this embodiment is a risk list generation device, which can be integrated into an electronic device, such as an electronic device related to a claims system. The risk list generation method provided in this embodiment includes the following steps:
[0082] Step S101, using the random forest bagging algorithm to generate a corresponding risk prediction model based on a preset data set; the preset data set includes historical claims related data; the historical claims related data includes: customer information, claims information and policy information.
[0083] In this embodiment, bagging is an integrated learning method based on the self-service sampling method, which improves the stability and generalization ability of the model by building multiple base learners and combining their predictions. On this basis, random forest introduces random attribute selection to enhance diversity. In the decision tree, random forest evaluates model performance through feature importance calculation and out-of-bag estimation. This method shows advantages in reducing overfitting and improving accuracy, and the computational overhead is relatively small.
[0084] The preset data set can be pre-set and used to train and generate the risk prediction model, including historical claims related data. The historical claims related data includes: customer information, claims information and policy information. Customer information can further include information such as customer name, customer category, customer age, etc. The claims information can further include: claims amount, cause of accident, reporting time, claims verification time, case closing time, etc. The policy information can include terms and conditions, policy category, policy signing time, etc.
[0085] Step S102, obtaining target claims data to be processed.
[0086] In this embodiment, the acquisition method may be to directly acquire from other related electronic devices, or to acquire raw data from other related electronic devices and then generate the data after data preprocessing.
[0087] Step S103: input the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data.
[0088] In this embodiment, the target claim data is input into the risk prediction model to generate a corresponding prediction result, that is, a risk label corresponding to the target claim data, which provides a basis for the subsequent generation of a risk list.
[0089] Step S104: generating a risk list corresponding to the target claim data according to the risk tag.
[0090] In this embodiment, risk labels may be pre-set, for example, labels corresponding to risk types, risk levels, etc. may be set, wherein labels of a single dimension may be set, or labels combining multiple dimensions may be set.
[0091] The risk list can include customer information, claims-related information, and risk information. At the same time, a template can be pre-set for the risk list, so that the claims data can be directly imported into the template, improving the efficiency of subsequent risk list generation.
[0092] The risk list generation method of this embodiment adopts the random forest bagging algorithm to generate a corresponding risk prediction model based on a preset data set; the preset data set includes historical claims related data; the historical claims related data includes: customer information, claims information and policy information; obtaining the target claim data to be processed; inputting the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data; and generating a risk list corresponding to the target claim data according to the risk label.
[0093] The risk list generation method of the present application generates a corresponding risk prediction model based on a preset data set through a random forest bagging algorithm, so that when processing target claims data, a risk label corresponding to the target claims data can be generated based on the risk prediction model. At the same time, a risk list corresponding to the target claims data is generated based on the risk label, so that the data risk situation and the specific risk list in the claims data can be determined proactively and more efficiently.
[0094] Figure 3 A flow chart of a method for generating a risk list provided in another embodiment of the present application is shown as follows: Figure 3 As shown, the risk list generation method provided in this embodiment is a further refinement of the risk list generation method provided in the previous embodiment of the present application. The risk list generation method provided in this embodiment includes the following steps.
[0095] Step S201, performing feature extraction processing on a preset data set to generate a corresponding feature set.
[0096] In this embodiment, the preset data set to be processed can be determined in advance, and the judgment dimension of the prediction result required to build the risk prediction model can be determined. The preset data set can be historical claims data, and the judgment dimension can be the compensation amount dimension, the number of compensation dimensions, the customer dimension, the claims operation dimension of the claims case handler, etc.
[0097] At the same time, the features extracted can be set according to the risk correlation. For example, when the compensation amount is large, the risk may be large, so the feature extracted can be set as the compensation amount, and other features can also be set accordingly according to the actual application. At the same time, optionally, it can also be set to be the same as the judgment dimension.
[0098] Step S202: Use the random forest bagging algorithm to construct multiple decision trees based on the feature set.
[0099] In this embodiment, the specific construction process may be as follows:
[0100] According to the preset conditions, such as features with greater risk association, first extract the feature set that meets the conditions to build a decision tree. Combine the features in the continuously input data to determine which decision tree to add. Each decision tree has gini (average weighted gini), samples (the number of observations in the node), value (the number of samples in each class), and class (the classification of most points in the node). As the number of samples increases, the smaller the gini of the result fed back by the decision tree, the more accurate the result. The specific generation rules are as follows:
[0101] A. Filter the training set in the database, randomly place and replace n training samples from the training set, and repeat X times;
[0102] B. If the feature dimension is O, specify a constant smaller than o, randomly select o feature subsets from O features, and select the best one from these o features each time the decision tree splits;
[0103] C. The generated X decision trees are combined into a random forest to ensure that each tree grows to the maximum extent possible without data loss;
[0104] D. Generate classification results based on X decision trees, and use the average of the prediction results of X trees as the final prediction result. Among them, the above n, X, O, and o can be set according to the actual application.
[0105] Step S203, constructing a corresponding risk prediction model based on the multiple decision trees finally constructed.
[0106] In this embodiment, the rule engine can be further combined to build an association between the decision tree prediction results and the risk labels, so as to ultimately generate the required risk prediction model.
[0107] Optionally, in this embodiment, S203 may be specifically as follows:
[0108] A preset rule engine is used to configure the mapping relationship between the risk dimension prediction results generated by multiple decision trees and the risk labels to generate a corresponding risk prediction model; each risk dimension in the risk dimension prediction results is a pre-set dimensional data.
[0109] In this embodiment, the dimension data may be a compensation amount dimension, a compensation number dimension, a customer dimension, a claim operation dimension of a claim case handler, etc. The preset rule engine may adopt a commonly used rule engine application, which is not limited in this embodiment.
[0110] Optionally, Flik stream processing technology can be used to combine multiple decision trees with rule engines to generate risk labels for cases. The use of Flik stream processing technology can reduce the complexity of building and arranging risk prediction models and improve building efficiency.
[0111] Optionally, in this embodiment, after the corresponding risk prediction model is generated, the risk prediction model may be further tested:
[0112] Obtaining a test sample; the test sample includes test claims data;
[0113] Inputting the test sample into a risk prediction model to generate a test risk label corresponding to the test sample;
[0114] Compare the test risk label with the actual risk label; the actual risk label is the risk label manually annotated on the test claim data;
[0115] If the test risk label is the same as the actual risk label, the test result of the risk prediction model is determined to be a passed test.
[0116] In this embodiment, the test claim data can be simulated data or historical claim data, which is different from the historical claim data in the above embodiment, and is used to test the accuracy of the risk prediction model. The actual risk label can be a risk label that is manually pre-labeled on the test claim data.
[0117] Optionally, the test risk label and the actual risk label may be completely identical or partially different. That is, if the number of differences between the test risk label and the actual risk label is low, for example, if the ratio of the different numbers is less than a preset ratio threshold, the test risk label can be considered to be the same as the actual risk label.
[0118] Step S204, obtaining target claim data to be processed.
[0119] Optionally, in this embodiment, the S204 may be specifically as follows:
[0120] Sending a claim data acquisition request to a claim-related electronic device;
[0121] Receiving corresponding initial claim data sent by electronic equipment related to claims;
[0122] The initial claims data is cleaned and preprocessed to generate target claims data.
[0123] In this embodiment, the claim data acquisition request may include the claim data identifier or claim data range information to be acquired, etc. Data cleaning and data preprocessing can be used to clean erroneous data, unify the data format, etc., to facilitate the generation of risk labels corresponding to subsequent target claim data.
[0124] Step S205, input the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data.
[0125] Optionally, in this embodiment, the risk prediction model may be further optimized in practical applications, as follows:
[0126] Obtaining misjudgment list information from a preset database; the misjudgment list information is claim data that is not on the risk list and has risks or claim data that is on the risk list but has no risks;
[0127] The risk prediction model is retrained based on the misjudgment list information to generate an optimized risk prediction model.
[0128] In this embodiment, the misjudged list information is the list information that the risk prediction model misjudged, such as the claims data that have risks but are not generated in the risk list, and the claims data that do not have risks but are in the risk list. The misjudged list information can be generated by manual review.
[0129] By retraining the risk prediction model with the information of the misjudgment list, the risk prediction model can be continuously optimized, thereby improving the accuracy of the risk prediction model.
[0130] Step S206, determining the associated data corresponding to the risk tag in the target claim data.
[0131] In this embodiment, the risk tag is associated with certain data in the target claim data, and the certain data is the above-mentioned associated data.
[0132] Step S207, importing the associated data into a preset list template to generate a corresponding risk list.
[0133] In this embodiment, the preset list template may be a commonly used Excel template or other preset templates, which is not limited in this embodiment.
[0134] In order to further illustrate the risk list generation method of this embodiment in detail, it will be further explained below in combination with specific applications.
[0135] The implementation of the risk list generation method of this embodiment mainly focuses on the following four key steps:
[0136] Step 1: Determine the needs analysis
[0137] In this step 1, the requirements are sorted out, and data exploration is performed on the customer information, claim information, and claim policy information in the production database. Combined with the production data, a preliminary claims risk prediction model framework and risk list generation dimensions are developed to determine the risk types and risk sources that need to be monitored.
[0138] Step 2: Configure risk dimensions
[0139] Based on the results of data exploration in step 1, configure risk monitoring dimensions for multiple insurance categories.
[0140] (1) Generation of risk control rules: For example, for claims involving more than three loss estimate adjustments, pay attention to the claims settlement operations of the case handler.
[0141] (2) Risk list generation: For example, for major claims with an estimated loss of more than 1 million yuan, we focus on the claims of the policyholders, the insured, and the payees in the past 10 years. The risk detection dimensions include the amount of compensation and the type of payee.
[0142] Step 3: Risk control rule model and risk control rule generation
[0143] In this step 3, the following designs are to be implemented:
[0144] (1) Through the random forest bagging algorithm, the risk prediction model is generated using customer information, claim information, policy information, expert knowledge base and other data in the database as the data set.
[0145] (2) According to the preset conditions, first extract the feature set that meets the conditions to build a decision tree. Combine the features in the continuously input data to determine which decision tree to add. Each decision tree has gini (average weighted gini), samples (the number of observations in the node), value (the number of samples in each class), and class (the classification of most points in the node). As the number of samples increases, the smaller the gini of the result fed back by the decision tree, the more accurate the result. The specific generation rules are as follows:
[0146] A. Filter the training set in the database, randomly place and replace n training samples from the training set, and repeat X times;
[0147] B. If the feature dimension is O, specify a constant smaller than o, randomly select o feature subsets from O features, and select the best one from these o features each time the decision tree splits;
[0148] C. The generated X decision trees are combined into a random forest to ensure that each tree grows to the maximum extent possible without data loss;
[0149] D. Generate classification results based on X decision trees, and use the mean of the prediction results of X trees as the final prediction result.
[0150] (3) Use Flik stream processing technology, combined with the imperfect risk prediction model, namely the decision tree and rule engine generated by the random forest bagging algorithm mentioned above, to generate risk labels for cases.
[0151] (4) In the case handling process of the non-auto claims system, the risk prediction model is combined with certain risk warnings. For risk labels of different degrees, different rules in the rule engine are executed to determine whether the case is allowed to be compensated or whether the compensation amount is reasonable.
[0152] (5) Test verification: After the rule configuration is completed, use simulated data or historical data for test verification to ensure the correctness and effectiveness of the risk rules and risk list generation.
[0153] Step 4: Configure the rules engine
[0154] In this step 4, the following designs are to be implemented:
[0155] (1) Select a rule engine: Select a suitable rule engine platform that provides a visual rule editor and powerful rule processing capabilities.
[0156] (2) Define rule configuration strategy: Define rule configuration strategy in the rule engine and configure rules according to the rule configuration strategy.
[0157] Among them, the rule configuration strategy includes the generation standard of risk labels, so the configured rules usually include the mapping relationship between conditions (i.e., conditions of risk dimensions) and risk labels. Conditions can be a combination of one or more dimensions. Risk labels can include identifiers indicating risk level, risk type, risk source, etc.
[0158] Optionally, the present application may set priorities. If multiple rules are satisfied at the same time, corresponding priorities need to be set for different rules to determine the final risk label.
[0159] Step 5: Apply the configured rules.
[0160] (1) Obtain the data to be tested.
[0161] Data source access: Connect the data sources that need to be monitored to the system. These data sources may include databases, API (Application Program Interface) interfaces, files, etc.
[0162] (2) Data cleaning: Clean and preprocess the data to be tested obtained from the data source to obtain processed data.
[0163] (3) Automatic label generation: The improved risk prediction model is used to match the rules of the processed data. If any one or more rules are matched, a risk label corresponding to the rule is generated.
[0164] In this embodiment, an expert knowledge base is introduced, and the expert knowledge base is connected to the improved risk prediction model.
[0165] The improved risk prediction model includes a parameter control module and a risk prediction model. The parameter control module is connected to the expert knowledge base, obtains information on risk false alarm / omission cases from the expert knowledge base, and tunes the parameters of the risk prediction model based on the information on risk false alarm / omission cases to improve the accuracy of risk list data generation.
[0166] In an optional implementation, for real-time data sources, a stream processing framework can be used to perform real-time data processing and rule matching to generate risk labels. In another optional implementation, for batch data sources (such as historical data in a database), batch processing tasks can be performed regularly to perform rule matching on the data and generate risk labels.
[0167] Optionally, the generated risk tags are stored in a database and associated with the original data to be detected to facilitate subsequent query and analysis.
[0168] (4) Determine the corresponding action based on the risk label
[0169] It should be understood that in the face of different risk labels, this embodiment may subsequently perform different actions on the data that meets the rule (ie, operations performed after the conditions are met).
[0170] In this embodiment, through preliminary data exploration and random forest bagging algorithm, a risk prediction model is developed using production data as a data set, and risk control rules are automatically generated through the model to provide risk warning information for each process and link of case handling.
[0171] By configuring the risk list rule engine and setting dimension conditions, risk label data can be automatically generated. This technical solution realizes the automatic identification and labeling of risks, improving the efficiency and accuracy of risk management.
[0172] At the same time, by introducing the expert knowledge base, the accuracy of the risk prediction model can be significantly improved, false positives and omissions can be reduced, and more reliable and effective support can be provided for risk management.
[0173] Optionally, in this embodiment, the following optimization may also be performed:
[0174] First point, monitoring and optimization
[0175] Monitoring: Monitor the model's running results by including indicators such as rule matching success rate, processing time, and error rate.
[0176] Logging: Records the log information of system operation, including rule matching results, exception information, etc., to facilitate problem troubleshooting and auditing.
[0177] Optimization: Based on monitoring results and business needs, optimize and adjust the rule engine to improve system performance and accuracy.
[0178] Second point: safety and compliance testing
[0179] Permission management: Perform permission control on the access and operation of the rule engine to ensure that only authorized personnel can modify the rule configuration.
[0180] Data encryption: Encrypt sensitive data for storage and transmission to ensure data security.
[0181] Compliance checks: Ensure that the system operates in compliance with relevant regulations and policy requirements, such as data protection regulations.
[0182] Figure 4 A schematic diagram of the structure of the risk list generation device provided in this application, such as Figure 4 As shown, in this embodiment, the risk list generating device 300 includes:
[0183] The first generating module 301 is used to generate a corresponding risk prediction model based on a preset data set using a random forest bagging algorithm; the preset data set includes historical claims-related data; the historical claims-related data includes: customer information, claims information and policy information;
[0184] An acquisition module 302 is used to acquire target claims data to be processed;
[0185] An input module 303 is used to input the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data;
[0186] The second generating module 304 is used to generate a risk list corresponding to the target claim data according to the risk tag.
[0187] The risk list generation device provided in this embodiment can execute Figure 2 The technical solution of the method embodiment shown in the figure has the same implementation principle and technical effect as Figure 2 The method embodiments shown are similar and will not be described in detail here.
[0188] The risk list generating device provided in the present application is further refined on the basis of the risk list generating device provided in the previous embodiment, and the risk list generating device 300 includes:
[0189] Optionally, in this embodiment, the first generating module 301 is specifically used for:
[0190] Perform feature extraction on the preset data set to generate a corresponding feature set; use the random forest bagging algorithm to construct corresponding multiple decision trees based on the feature set; and construct a corresponding risk prediction model based on the multiple decision trees finally constructed.
[0191] Optionally, in this embodiment, when the first generating module 301 constructs a corresponding risk prediction model based on the multiple decision trees finally constructed, it is specifically used to:
[0192] A preset rule engine is used to configure the mapping relationship between the risk dimension prediction results generated by multiple decision trees and the risk labels to generate a corresponding risk prediction model; each risk dimension in the risk dimension prediction results is a pre-set dimensional data.
[0193] Optionally, in this embodiment, the acquisition module 302 is specifically used for:
[0194] Sending a claim data acquisition request to an electronic device related to claim settlement; receiving corresponding initial claim data sent by the electronic device related to claim settlement; performing data cleaning and data preprocessing on the initial claim data to generate target claim data.
[0195] Optionally, in this embodiment, the second generating module 304 is specifically used to:
[0196] Determine the associated data corresponding to the risk tag in the target claims data; import the associated data into a preset list template to generate a corresponding risk list.
[0197] Optionally, in this embodiment, the risk list generating device 300 further includes:
[0198] A training module is used to obtain misjudgment list information from a preset database; the misjudgment list information is claim data that is not on the risk list and has risks, or claim data that is on the risk list but does not have risks; based on the misjudgment list information, the risk prediction model is re-trained to generate an optimized risk prediction model.
[0199] Optionally, in this embodiment, the first generating module 301 is further used to:
[0200] Obtain a test sample; the test sample includes test claims data; input the test sample into a risk prediction model to generate a test risk label corresponding to the test sample; compare the test risk label with the actual risk label; the actual risk label is a risk label manually annotated on the test claims data; if the test risk label is the same as the actual risk label, determine that the test result of the risk prediction model is a passed test.
[0201] The risk list generation device provided in this embodiment can execute Figure 2-Figure 3 The technical solution of the method embodiment shown in the figure has the same implementation principle and technical effect as Figure 2-Figure 3 The method embodiments shown are similar and will not be described in detail here.
[0202] According to an embodiment of the present application, the present application also provides an electronic device, a computer-readable storage medium and a computer program product.
[0203] like Figure 5 As shown, Figure 5 is a schematic diagram of the structure of the electronic device provided by the present application. The electronic device is intended for various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, blade servers, mainframe computers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.
[0204] like Figure 5 As shown, the electronic device includes: a processor 401 and a memory 402. The various components are connected to each other using different buses and can be installed on a common mainboard or in other ways as needed. The processor can process instructions executed in the electronic device. The electronic device can also include a transceiver, which is used to send and receive data.
[0205] The memory 402 is a non-transitory computer-readable storage medium provided in the present application. The memory stores instructions executable by at least one processor to enable at least one processor to execute the risk list generation method provided in the present application. The non-transitory computer-readable storage medium of the present application stores computer instructions, which are used to enable a computer to execute the risk list generation method provided in the present application.
[0206] The memory 402 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the risk list generation method in the embodiment of the present application (for example, the attached Figure 4 The processor 401 executes various functional applications and data processing of the electronic device by running the non-transient software programs, instructions and modules stored in the memory 402, that is, the risk list generation method in the above method embodiment is implemented.
[0207] This embodiment also provides a computer-readable storage medium. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute the risk list generation method of the above embodiment.
[0208] At the same time, this embodiment also provides a computer product. When the instructions in the computer product are executed by a processor of an electronic device, the electronic device can execute the risk list generation method of the above embodiment.
[0209] Those skilled in the art will readily come up with other implementations of the embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modifications, uses or adaptations of the embodiments of the present application, which follow the general principles of the embodiments of the present application and include common knowledge or customary technical means in the art that are not disclosed in the embodiments of the present application.
[0210] It should be understood that the embodiments of the present application are not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the embodiments of the present application is limited only by the appended claims.
Claims
1. A method for generating a risk list, characterized in that: include: The random forest bagging algorithm is used to generate the corresponding risk prediction model based on the preset data set; The preset data set includes historical claims-related data; The historical claims-related data include: customer information, claims information and policy information; Obtain target claims data to be processed; Inputting the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data; A risk list corresponding to the target claims data is generated according to the risk tags.
2. The method according to claim 1, characterized in that: The random forest bagging algorithm is used to generate a corresponding risk prediction model based on a preset data set, including: Perform feature extraction on the preset data set to generate a corresponding feature set; A random forest bagging algorithm is used to construct a corresponding plurality of decision trees based on the feature set; The corresponding risk prediction model is constructed based on the multiple decision trees finally constructed.
3. The method according to claim 2, characterized in that The corresponding risk prediction model is constructed based on the multiple decision trees finally constructed, including: A preset rule engine is used to configure the mapping relationship between the risk dimension prediction results generated by multiple decision trees and the risk labels to generate a corresponding risk prediction model; each risk dimension in the risk dimension prediction results is a pre-set dimensional data.
4. The method according to claim 1, characterized in that: The obtaining of target claims data to be processed includes: Sending a claim data acquisition request to a claim-related electronic device; Receiving corresponding initial claim data sent by electronic equipment related to claims; The initial claims data is cleaned and preprocessed to generate target claims data.
5. The method according to claim 1, characterized in that The step of generating a risk list corresponding to the target claim data according to the risk tag includes: Determining associated data corresponding to the risk tag in the target claims data; Import the associated data into the preset list template to generate the corresponding risk list.
6. The method according to claim 1, characterized in that The method further comprises: Obtaining misjudgment list information from a preset database; the misjudgment list information is claim data that is not on the risk list and has risks or claim data that is on the risk list but has no risks; The risk prediction model is retrained based on the misjudgment list information to generate an optimized risk prediction model.
7. The method according to claim 3, characterized in that After the mapping relationship between the risk dimension prediction results generated by the plurality of decision trees and the risk labels is configured by using the preset rule engine to generate the corresponding risk prediction model, the method further includes: Obtaining a test sample; the test sample includes test claims data; Inputting the test sample into a risk prediction model to generate a test risk label corresponding to the test sample; Compare the test risk label with the actual risk label; the actual risk label is the risk label manually annotated on the test claim data; If the test risk label is the same as the actual risk label, the test result of the risk prediction model is determined to be a passed test.
8. A risk list generating device, characterized in that: include: The first generation module is used to generate a corresponding risk prediction model based on a preset data set using a random forest bagging algorithm; The preset data set includes historical claims-related data; The historical claims-related data include: customer information, claims information and policy information; An acquisition module is used to acquire target claims data to be processed; An input module, used for inputting the target claim data into the risk prediction model to generate a risk label corresponding to the target claim data; The second generating module is used to generate a risk list corresponding to the target claim data according to the risk label.
9. An electronic device, characterized in that: include: Memory and processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the risk list generation method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the risk list generation method as described in any one of claims 1 to 7.
11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the risk list generation method according to any one of claims 1 to 7 is implemented.