Method and system for identifying insurance fraud, electronic equipment and storage medium

Abstract

The invention discloses a method and a system for identifying insurance fraud, electronic equipment and a storage medium, and the method for identifying insurance fraud specifically comprises the following steps: 1, obtaining insurance policy information of corresponding claims settlement persons in insurance policies of the same type in an insurance claim settlement database, and carrying out the classification according to feature vectors; step 2, performing data cleaning on the classified insurance claim insurance policy information, and performing feature extraction processing on the cleaned data; 3, acquiring external data associated with the feature data, performing association analysis verification according to the data analysis verification model, and pre-selecting a fraud factor; and step 4, on the basis of the height of the fraud risk value, the pre-selected fraud factor with the high fraud risk value and the pre-selected fraud factor with the low fraud risk value are divided into two groups, namely a first fraud factor and a second fraud factor. According to the invention, through efficient data processing and analysis and accurate fraud factor identification, the accuracy of fraud identification is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, system, electronic device and storage medium for identifying insurance fraud. Background Art

[0002] Insurance fraud refers to the act of defrauding insurance money from insurance companies by means of fictitious insurance subject matter, fabricating false accident causes, exaggerating the extent of losses, etc. for the purpose of defrauding insurance money. Insurance fraud not only damages the interests of insurance companies, but also destroys the fair order of the insurance market and increases the insurance costs of other policyholders. With the rapid development of the insurance industry, insurance fraud has become increasingly serious, causing huge economic losses to insurance companies and society.

[0003] Patent (CN114723574A) is an insurance anti-fraud identification method, device, equipment and medium based on artificial intelligence, which proposes to directly determine the suspected fraud gang members by traversing the updated network relationship through the target edge, reducing the number of traversals of target nodes, edges and paths, reducing the traversal time, and improving the traversal efficiency; and patent (CN109801176B) is a method, system, electronic device and storage medium for identifying insurance fraud, which determines whether the insurance data to be identified is fraudulent by applying the second insurance fraud monitoring model. However, in the prior art, there is no connection relationship between various data, there is a problem of inaccurate judgment, and the success rate of monitored insurance fraud is low and the false alarm rate is high, which has caused great losses to insurance companies. Summary of the invention

[0004] In view of the defects in the prior art, the present invention provides a method for identifying insurance fraud, which specifically includes the following steps:

[0005] Step 1. Obtain the insurance policy information of the corresponding claimant in the insurance policy of the same type in the insurance claim database, and classify them according to the feature vector;

[0006] Step 2: Clean the classified insurance claim policy information and perform feature extraction on the cleaned data;

[0007] Step 3. Obtain external data associated with the feature data, perform correlation analysis verification based on the data analysis verification model, and pre-select fraud factors;

[0008] Step 4. Based on the height of the fraud risk value, the pre-selected fraud factors with high fraud risk values ​​and the pre-selected fraud factors with low fraud risk values ​​are divided into two groups, namely, first fraud factors and second fraud factors;

[0009] Step 5. Use the training set containing the first fraud factor to input into a neural network-based support vector machine model to generate a first fraud assessment model;

[0010] Using the training set containing the second fraud factor to input into the support vector machine model based on the neural network to generate a second fraud assessment model; and using the test set to verify and optimize the fraud assessment model;

[0011] Step 6. Based on the first fraud assessment model and the second fraud assessment model, the first fraud factor and the second fraud factor are evaluated respectively to verify whether the fraud factor exists and determine whether there is insurance claim fraud.

[0012] Preferably, a preliminary machine learning model is established based on the extracted feature data to detect fraud, including using cross-validation techniques to evaluate the effectiveness of the model, and using different features of external data to optimize the model and observe the improved performance.

[0013] Preferably, in step 3, the external data and the feature data need to be integrated to ensure that their formats and timestamps are consistent.

[0014] Preferably, in step 5, high risk factors include short account history, wrong guarantor and wrong specific information; low risk factors include local deviation of accident location, deviation of maintenance information and wrong information.

[0015] Preferably, the system for identifying insurance fraud includes a fraud identification system, wherein the fraud identification system has the following modules:

[0016] Data acquisition module, responsible for collecting data from various sources;

[0017] Data analysis module, which performs preliminary analysis and processing on the collected data;

[0018] Feature extraction module, which extracts feature values ​​from raw data that are helpful for subsequent analysis and modeling;

[0019] Data pre-selection module, which performs intelligent classification and pre-selection based on preset feature value data;

[0020] Intelligent classification module, which uses deep learning algorithms to automatically classify data;

[0021] Preferably, the fraud identification system further comprises an intelligent fraud assessment model, the intelligent fraud assessment model comprises a first fraud assessment model and a second fraud assessment model, and the intelligent fraud assessment model specifically comprises the following functional modules:

[0022] A training module, used to use a data set including fraud factors as training samples, and use a convolutional neural network to train a fraud assessment model on a number of the training samples to generate an intelligent fraud assessment model;

[0023] The verification module is used to verify the intelligent fraud assessment model with the test set and complete the verification and optimization of the model;

[0024] A construction module, for inputting a training set containing fraud factors into a support vector machine model based on a neural network to construct a fraud assessment model;

[0025] The judgment module is applied to the intelligent fraud assessment model to determine whether the fraud factor to be identified is fraudulent.

[0026] Preferably, the data acquisition module is connected to the data analysis module, and the data analysis module is used to analyze and process the data collected by the data acquisition module; the data pre-selection module performs pre-selection and classification based on the feature values ​​extracted by the feature extraction module.

[0027] Preferably, an electronic device comprises an intelligent processing module and an intelligent storage module, wherein the intelligent storage module stores computer-readable instructions, and the computer-readable instructions implement the method according to any one of claims 1-2 when executed by the intelligent processing module.

[0028] Preferably, a computer-readable storage medium stores a computer program, and the computer program executes the method according to any one of claims 1-2 when executed by a processor or a computer.

[0029] The beneficial effects of the present invention are embodied in:

[0030] The present invention can accurately identify key factors closely related to insurance fraud by verifying and pre-selecting fraud factors through association analysis, and further improve the accuracy of fraud identification by grouping factors with high fraud risk values ​​and factors with low fraud risk values; at the same time, through efficient data processing and analysis, accurate fraud factor identification, powerful model building and optimization capabilities, flexible system architecture and scalability, and wide applicability and compatibility, it can automatically extract valuable information from massive data, reduce manual intervention, and reduce error rates, which not only improves the accuracy and efficiency of insurance fraud identification, but also provides strong technical support and guarantee for related industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0032] Figure 1 It is a schematic diagram of the system framework of the present invention;

[0033] Figure 2 It is a structural function diagram of the intelligent fraud assessment model of the present invention;

[0034] Figure 3 It is a structural schematic diagram of the electronic device of the present invention;

[0035] Figure 4 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0036] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0037] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.

[0038] like Figure 4 As shown,

[0039] The method of identifying insurance fraud includes the following steps:

[0040] Step 1. Obtain the insurance policy information of the corresponding claimant in the insurance policy of the same type in the insurance claim database, and classify them according to the feature vector;

[0041] Step 2: Clean the classified insurance claim policy information and perform feature extraction on the cleaned data;

[0042] Step 3. Obtain external data associated with the feature data, perform correlation analysis verification based on the data analysis verification model, and pre-select fraud factors;

[0043] Step 4. Based on the height of the fraud risk value, the pre-selected fraud factors with high fraud risk values ​​and the pre-selected fraud factors with low fraud risk values ​​are divided into two groups, namely, first fraud factors and second fraud factors;

[0044] Step 5. Use the training set containing the first fraud factor to input into a neural network-based support vector machine model to generate a first fraud assessment model;

[0045] Using the training set containing the second fraud factor to input into the support vector machine model based on the neural network to generate a second fraud assessment model; and using the test set to verify and optimize the fraud assessment model;

[0046] Step 6. Based on the first fraud assessment model and the second fraud assessment model, the first fraud factor and the second fraud factor are evaluated respectively to verify whether the fraud factor exists and whether there is insurance claim fraud;

[0047] Specifically, the same type of insurance policies in step 1 refer to insurance policies with the same insurance product code and similar coverage. They are classified by using the time point as the feature vector, that is, the specific information of "same type of insurance policies" and "corresponding claims adjusters" is clarified, and the query is executed in the database through API calls using a programming language, and the returned data is sorted to complete statistical analysis and data mining;

[0048] The data feature extraction in step 2 includes the claimant's name, policy type, policy signing time, claim activation time and specific location in the data as feature values ​​for feature extraction, that is, the initial insurance data is cleaned, deduplicated and integrated to generate an insurance data set, and the fields of the initial insurance data are analyzed to obtain the field attribute types of the initial insurance data and classify them. According to the pre-set feature values, the classified and extracted initial insurance data is uniformly converted into standard data feature values;

[0049] The data analysis and verification model in step 3 analyzes the data using the logistic regression method, uses the Apriori algorithm to discover association rules in the data, and finally uses the isolation forest method to identify abnormal data and select pre-selected fraud factors; the external data includes medical records, vehicle maintenance records, public databases, etc., that is, the data analysis and verification model will perform correlation analysis on the extracted feature data and external data to ensure that their formats and timestamps are consistent, and verify the authenticity of the claim information, including comparing the medical diagnosis provided by the insured with the actual records of the hospital to see if there are any differences;

[0050] At the same time, the selection of pre-selected fraud factors is based on the characteristic data that have differences after the data analysis verification model is verified.

[0051] For the fraud factor grouping in step 4, the grouping is divided by the standard threshold, and the standard threshold is divided by the direct difference and deviation difference in the pre-selected fraud factors. For example, an error in the name of the insured is a direct difference, and an error in the maintenance record or the accident location is a deviation difference, that is, the direct difference is judged as a high fraud risk value, and the deviation difference is judged as a low fraud risk value; the fraud risk value of each observation is calculated using a logistic regression model, and these risk values ​​are expressed as probabilities or scores. The higher the score, the greater the fraud risk; for example, the threshold is set, high risk factor (first fraud factor): fraud risk value>0.5; low risk factor (second fraud factor): fraud risk value≤0.5;

[0052] For steps 5 and 6:

[0053] Input the training set to ensure that the training data sets of the first and second fraud factors are cleaned and preprocessed, and perform feature selection to select features related to fraud assessment and perform standardization; then use the training set containing the first fraud factor to input the first neural network-based support vector machine (NN-SVM) model to train the model, perform validation analysis in a cross-validation manner, and adjust hyperparameters to optimize model performance; at the same time, input the training set containing the second fraud factor into the second NN-SVM model, and perform validation analysis and optimization in the same cross-validation manner;

[0054] The first fraud factor data is predicted by using a first fraud assessment model to identify possible first fraud factors. The second fraud factor is predicted by using a second fraud assessment model to identify possible second fraud factors. If the model predicts the existence of a fraud factor, further investigation and evidence collection is performed to confirm whether there is insurance claim fraud.

[0055] Embodiment 2

[0056] like Figure 1-2 The system for identifying insurance fraud includes a fraud identification system, and the fraud identification system includes the following modules:

[0057] The data acquisition module is responsible for collecting data from various sources, including databases, files, networks and other channels.

[0058] The data analysis module performs preliminary analysis and processing on the collected data; uses the NumPy data processing library to perform data cleaning (including removing duplicate, erroneous or irrelevant data), data conversion (such as standardization and normalization), and statistical analysis (calculating mean, median, variance, etc.);

[0059] Feature extraction module, which extracts feature values ​​from raw data that are helpful for subsequent analysis and modeling;

[0060] The data pre-selection module performs intelligent classification and pre-selection based on preset characteristic value data; pre-selection and screening are performed based on preset characteristic values ​​in a conditional filtering manner;

[0061] Intelligent classification module, which uses deep learning algorithms to automatically classify data;

[0062] The data acquisition module is connected to the data analysis module, and the data analysis module is used to analyze and process the data collected by the data acquisition module; the data pre-selection module performs pre-selection and classification based on the feature values ​​extracted by the feature extraction module; through the independent and close cooperation between these modules, a complete and flexible system architecture is formed, the system is easy to expand and maintain, and new functional modules can be added or existing modules can be optimized according to actual needs;

[0063] The fraud identification system also includes an intelligent fraud assessment model, which includes a first fraud assessment model and a second fraud assessment model. The intelligent fraud assessment model specifically includes the following functional modules:

[0064] A training module, used to use a data set including fraud factors as training samples, train a fraud assessment model on a number of training samples using a convolutional neural network, and generate an intelligent fraud assessment model;

[0065] The verification module is used to verify the intelligent fraud assessment model with the test set and complete the verification and optimization of the model;

[0066] A construction module, for inputting a training set containing fraud factors into a support vector machine model based on a neural network to construct a fraud assessment model;

[0067] The judgment module is applied to the intelligent fraud assessment model to determine whether the fraud factor to be identified is fraudulent.

[0068] Embodiment 3

[0069] like Figure 3 , an electronic device, wherein the electronic device includes but is not limited to a smart phone, a landline phone, a tablet computer, a laptop computer, a wearable device and other electronic devices, wherein an intelligent processing module and an intelligent storage module are arranged inside the electronic device, and the intelligent storage module stores computer-readable instructions for executing the above-mentioned method for identifying insurance fraud of the present invention;

[0070] Embodiment 4

[0071] A computer-readable storage medium may be a ROM, an optical storage medium, a magnetic storage medium or other types of program storage. A computer program is stored on the computer-readable storage medium. When the computer program is run by a processor or a computer, the method for identifying insurance fraud of the present invention is executed.

[0072] That is, the present invention can be widely applied to various scenarios and devices by combining with electronic devices (such as computers, smart phones, etc.) and computer-readable storage media.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for identifying insurance fraud, characterized in that, The specific steps include: Step 1. Obtain the insurance policy information of the corresponding claimant in the insurance policy of the same type in the insurance claim database, and classify them according to the feature vector; Step 2: Clean the classified insurance claim policy information and perform feature extraction on the cleaned data; Step 3. Obtain external data associated with the feature data, perform correlation analysis verification based on the data analysis verification model, and pre-select fraud factors; Step 4. Based on the height of the fraud risk value, the pre-selected fraud factors with high fraud risk values ​​and the pre-selected fraud factors with low fraud risk values ​​are divided into two groups, namely, first fraud factors and second fraud factors; Step 5. Use the training set containing the first fraud factor to input into a neural network-based support vector machine model to generate a first fraud assessment model; Using the training set containing the second fraud factor to input into the support vector machine model based on the neural network to generate a second fraud assessment model; and using the test set to verify and optimize the fraud assessment model; Step 6: Based on the first fraud assessment model and the second fraud assessment model, the first fraud factor and the second fraud factor are evaluated respectively to verify whether the fraud factor exists and determine whether there is insurance claim fraud.

2. The method for identifying insurance fraud according to claim 1, characterized in that: A preliminary machine learning model is built based on the extracted feature data to detect fraud, including using cross-validation techniques to evaluate the effectiveness of the model, and using different features of external data to optimize the model and observe the improved performance.

3. The method for identifying insurance fraud according to claim 1, characterized in that: In step 3, the external data needs to be integrated with the feature data to ensure that their formats and timestamps are consistent.

4. The method for identifying insurance fraud according to claim 1, characterized in that: In step 5, high-risk factors include short account history, incorrect guarantor, and specific information errors; low-risk factors include local deviation of the accident location, maintenance information deviation and information errors.

5. A system for identifying insurance fraud, comprising a fraud identification system, characterized in that: The fraud identification system comprises the following modules: Data acquisition module, responsible for collecting data from various sources; Data analysis module, which performs preliminary analysis and processing on the collected data; Feature extraction module, which extracts feature values ​​from raw data that are helpful for subsequent analysis and modeling; Data pre-selection module, which performs intelligent classification and pre-selection based on preset feature value data; Intelligent classification module, which uses deep learning algorithms to automatically classify data.

6. The system for identifying insurance fraud according to claim 5, characterized in that: The fraud identification system further includes an intelligent fraud assessment model, and the intelligent fraud assessment model includes a first fraud assessment model and a second fraud assessment model.

7. The system for identifying insurance fraud according to claim 6, characterized in that: The intelligent fraud assessment model specifically includes the following functional modules: A training module, used to use a data set including fraud factors as training samples, and use a convolutional neural network to train a fraud assessment model on a number of the training samples to generate an intelligent fraud assessment model; The verification module is used to verify the intelligent fraud assessment model with the test set and complete the verification and optimization of the model; A construction module, for inputting a training set containing fraud factors into a support vector machine model based on a neural network to construct a fraud assessment model; The judgment module is applied to the intelligent fraud assessment model to determine whether the fraud factor to be identified is fraudulent.

8. The system for identifying insurance fraud according to claim 7, characterized in that: The data acquisition module is connected to the data analysis module, and the data analysis module is used to analyze and process the data collected by the data acquisition module; the data pre-selection module performs pre-selection and classification based on the feature values ​​extracted by the feature extraction module.

9. An electronic device, characterized in that: It comprises an intelligent processing module and an intelligent storage module, wherein the intelligent storage module stores computer-readable instructions, and the computer-readable instructions implement the method according to any one of claims 1-2 when executed by the intelligent processing module.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor or a computer, the method according to any one of claims 1 to 2 is executed.

Images