A smart claims settlement and abnormal transaction risk control system for the insurance industry

Abstract

This invention discloses an intelligent risk control system for abnormal transactions in the insurance industry, comprising a transaction sample acquisition subsystem for acquiring transaction sample data during the insurance claims process. This transaction sample data includes sample set data, basic account information data, fund transaction graphs, and transaction link data. A risk control subsystem is used to determine risk and risk control level based on the transaction sample data. The risk control subsystem includes a first risk control module for processing and determining risk control based on sample set data, a second risk control module for processing and determining risk control based on basic account information data, a third risk control module for processing and determining risk control based on fund transaction graphs, and a fourth risk control submodule for processing and determining risk control based on transaction work order data. This invention improves the accuracy of abnormal transaction risk control detection by having the risk control subsystem perform risk determination and risk control level assessment based on the above multi-dimensional data.

Description

Technical Field

[0001] This invention relates to the field of insurance claims, specifically to an intelligent risk control system for abnormal transactions in the insurance industry. Background Technology

[0002] With the continuous development of society and the economy, people are increasingly aware of the importance of insurance. According to the contract, the insured pays premiums to the insurer, and the insurer compensates for losses caused by risks stipulated in the contract. Therefore, risk assessment for insurance business is extremely important for insurers. Traditionally, risk assessment for insurance claims is usually conducted manually by agents based on their past experience and specific scenarios. This relies heavily on the agent's personal judgment, is highly subjective, and often results in low accuracy. As competition in the insurance industry intensifies and customer service awareness increases, traditional claims services, which rely primarily on manual risk management, are inefficient and have a high rate of case leakage. They can no longer meet the needs of customers and insurance companies. Therefore, the demand for rapid, accurate, and differentiated claims services is becoming increasingly urgent.

[0003] With the development of machine learning, current machine learning can only calculate the corresponding risk coefficients in advance by using offline historical data of the same type of business, and then exchange data with the database through the application programming interface. This method cannot guarantee the accuracy and timeliness of risk data. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent risk control system for abnormal transactions in insurance claims, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent insurance industry claims abnormal transaction risk control system, comprising: a transaction sample acquisition subsystem, used to acquire transaction sample data in the insurance claims process, the transaction sample data including sample set data, account basic information data, fund transaction graph and transaction link data; and a risk control subsystem, used to determine risk and risk control level based on the transaction sample data.

[0006] Preferably, the risk control subsystem includes a first risk control submodule that processes and determines risk based on sample set data, a second risk control submodule that processes and determines risk based on basic account information data, a third risk control submodule that processes and determines risk based on fund transaction graphs, and a fourth risk control submodule that processes and determines risk based on transaction link data.

[0007] The sample set data includes historical abnormal transaction data and transaction work order data. The transaction sample acquisition subsystem has a preset abnormal transaction identification model. This model performs invalid data removal operations on historical abnormal transaction data and transaction work order data based on a preset data completeness requirement, obtaining valid sample data. Feature filtering is then performed on the valid sample data to obtain training sample data. The second risk control submodule can determine whether the current transaction matches the target user's user profile based on the target user tags in the account basic information data. If the current transaction does not match the target user's user profile, the target transaction is processed as an anomaly. When the current transaction passes the transaction parameter verification, the target account information to which the current transaction belongs is obtained, and the target consumption information corresponding to the target user is determined based on the target user tags in the target account information. The current transaction information is compared with the target consumption information to determine whether the current transaction matches the target user's user profile.

[0008] The third risk control submodule is used to acquire node representation data in the fund transaction graph. The fund transaction graph consists of nodes and edges. Nodes represent the account data corresponding to the node, and edges between two nodes represent the transaction flow between the two nodes. Based on the fund transaction graph, the third risk control submodule generates a node feature set, wherein the node feature set consists of at least one node feature vector, and each node feature vector corresponds to a node in the fund transaction graph. The third risk control submodule inputs each node feature vector into a pre-trained graph model to obtain the prediction result corresponding to each node. The prediction result is used to characterize whether the account corresponding to each node is an abnormal fund transaction account. The model is trained based on a graph different from the fund transaction graph. The fourth risk control submodule includes a construction unit, an aggregation unit, and an anomaly identification unit. The construction unit is used to construct a transaction network graph corresponding to the transaction account to be detected, with the transaction account to be detected as the central node and the transaction account A in the user device and the transaction account B in the merchant device as the outer nodes. The aggregation module is used to perform feature aggregation processing on the node features of each node in the transaction network graph and obtain the aggregated features corresponding to the transaction account to be detected. The anomaly identification module performs anomaly identification on the transaction account to be detected based on the aggregated features to determine whether the transaction account to be detected is an abnormal transaction account.

[0009] Preferably, the first risk control submodule acquires the above-mentioned training sample data and constructs an abnormal transaction identification model based on logistic regression. By inputting the basic account information data into the abnormal transaction identification model, the abnormal transaction identification result is obtained, and the first abnormal transaction identification result is output.

[0010] Preferably, when the fourth risk control submodule determines that the transaction account to be detected is an abnormal transaction account, it also determines multiple transaction links associated with the transaction account to be detected, extracts link features from the account features of the transaction account in each transaction link, and determines whether the link features are abnormal link features or normal link features.

[0011] Preferably, the risk control subsystem also includes a risk control evaluation submodule. The risk control evaluation submodule performs index evaluation based on the first abnormal transaction identification result, the abnormal transaction account determination result, and the user profile determination result. When the evaluation index is greater than or equal to the preset alarm threshold, the corresponding alarm prompt information is generated.

[0012] Compared with the prior art, the beneficial effects of the present invention are:

[0013] Based on the acquired sample set data, basic account information data, fund transaction graphs, and transaction link data, the risk control subsystem performs risk assessment and risk control level determination based on the multi-dimensional data, thereby improving the accuracy of abnormal transaction risk control detection. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0015] Figure 2 This is a schematic diagram of the processing flow of the first risk control submodule in an embodiment of the present invention;

[0016] Figure 3 This is a schematic diagram of the processing flow of the second risk control submodule in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1-3 The present invention provides a technical solution: an intelligent claims settlement abnormal transaction risk control system for the insurance industry, including a transaction sample acquisition subsystem for acquiring transaction sample data during the insurance claims process, the transaction sample data including sample set data, basic account information data, fund transaction graph and transaction link data; and a risk control subsystem for determining risk and risk control level based on the transaction sample data.

[0019] In this embodiment, the risk control subsystem includes a first risk control submodule that processes and determines risk based on sample set data, a second risk control submodule that processes and determines risk based on basic account information data, a third risk control submodule that processes and determines risk based on fund transaction graphs, and a fourth risk control submodule that processes and determines risk based on transaction link data.

[0020] The sample set data includes historical abnormal transaction data and transaction work order data. The transaction sample acquisition subsystem has a preset abnormal transaction identification model. This model performs invalid data removal operations on historical abnormal transaction data and transaction work order data based on a preset data completeness, obtaining valid sample data. Feature filtering is then performed on the valid sample data to obtain training sample data. The second risk control submodule can determine whether the current transaction matches the target user's user profile based on the target user tags in the account basic information data. If the current transaction does not match the target user's user profile, the target transaction is processed as an anomaly. When the current transaction passes the transaction parameter verification, the target account information to which the current transaction belongs is obtained, and the target consumption information corresponding to the target user is determined based on the target user tags in the target account information. The current transaction information is compared with the target consumption information to determine whether the current transaction matches the target user's user profile. The third risk control submodule is used to obtain node representation data in the fund transaction graph. The fund transaction graph consists of nodes and edges. Node representation corresponds to the account data of the node, and the edges between two nodes... The third risk control submodule, representing the transaction flow between two nodes, generates a node feature set based on the fund transaction graph. Each node feature set consists of at least one node feature vector, with each feature vector corresponding to a node in the fund transaction graph. The third risk control submodule inputs each node feature vector into a pre-trained graph model to obtain a prediction result for each node. This prediction result is used to characterize whether the account corresponding to each node is an abnormal fund transaction account. The graph model is trained based on a graph different from the fund transaction graph. The fourth risk control submodule includes a construction unit, an aggregation unit, and an anomaly identification unit. The construction unit is used to construct a transaction network graph corresponding to the transaction account to be detected, with the transaction account to be detected as the central node and transaction account A in the user's device and transaction account B in the merchant's device as outer nodes. The aggregation module performs feature aggregation processing on the node features of each node in the transaction network graph to obtain the aggregated features corresponding to the transaction account to be detected. The anomaly identification module performs anomaly identification on the transaction account to be detected based on the aggregated features to determine whether the transaction account to be detected is an abnormal transaction account.

[0021] In this embodiment, the first risk control submodule acquires the above-mentioned training sample data and constructs an abnormal transaction identification model based on logistic regression. By inputting the basic account information data into the abnormal transaction identification model, the abnormal transaction identification result is obtained, and the first abnormal transaction identification result is output.

[0022] In this embodiment, when the fourth risk control submodule determines that the transaction account to be detected is an abnormal transaction account, it also determines multiple transaction links associated with the transaction account to be detected, extracts link features from the account features of the transaction account in each transaction link, and determines whether the link features are abnormal link features or normal link features.

[0023] In this embodiment, the risk control subsystem also includes a risk control evaluation submodule. The risk control evaluation submodule performs index evaluation based on the first abnormal transaction identification result, the abnormal transaction account determination result, and the user profile determination result. When the evaluation index is greater than or equal to the preset alarm threshold, the corresponding alarm prompt information is generated.

[0024] In this embodiment, to process large amounts of transaction data in real time, the open-source stream processing platform Kafka cluster is used as the processing architecture. Kafka, originally developed by LinkedIn, is a distributed, partitioned, multi-replica, high-throughput distributed publish-subscribe messaging system. It's a distributed messaging system coordinated by the distributed application coordination service ZooKeeper. Its greatest feature is its ability to process large amounts of data in real time to meet various application scenarios. When applied to a regular server, it can process hundreds of thousands of messages per second and can be used for low-latency collection and transmission of large amounts of event and log data.

[0025] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent claims settlement and abnormal transaction risk control system for the insurance industry, characterized in that, include: The transaction sample acquisition subsystem is used to acquire transaction sample data during the insurance claims process. The transaction sample data includes sample set data, basic account information data, fund transaction graphs, and transaction link data. The risk control subsystem is used to determine risk and risk control level based on transaction sample data; The risk control subsystem includes a first risk control submodule that processes and determines risk based on sample set data, a second risk control submodule that processes and determines risk based on basic account information data, a third risk control submodule that processes and determines risk based on fund transaction graphs, and a fourth risk control submodule that processes and determines risk based on transaction link data. The third risk control submodule is used to acquire node representation data in the fund transaction graph. The fund transaction graph consists of nodes and edges. Nodes represent the account data corresponding to the node, and the edges between two nodes represent the transaction flow between the two nodes. Based on the fund transaction graph, the third risk control submodule generates a node feature set, wherein the node feature set consists of at least one node feature vector, and each node feature vector corresponds to a node in the fund transaction graph. The third risk control submodule inputs each node feature vector into a pre-trained graph model to obtain the prediction result corresponding to each node. The prediction result is used to characterize whether the account corresponding to each node is an abnormal fund transaction account. The graph model is trained based on a graph different from the fund transaction graph. The fourth risk control submodule includes a construction unit, an aggregation unit, and an anomaly identification unit. The construction unit is used to construct a transaction network graph corresponding to the transaction account to be detected, with the transaction account to be detected as the central node and transaction account A in the user device and transaction account B in the merchant device as the outer nodes. The aggregation unit is used to perform feature aggregation processing on the node features of each node in the transaction network graph to obtain the aggregated features corresponding to the transaction account to be detected. The anomaly identification module performs anomaly identification on the transaction account to be detected based on the aggregated features to determine whether the transaction account to be detected is an abnormal transaction account.

2. The intelligent claims settlement abnormal transaction risk control system for the insurance industry according to claim 1, characterized in that: The sample set data includes historical abnormal transaction data and transaction work order data. The transaction sample acquisition subsystem has a preset abnormal transaction identification model. The abnormal transaction identification model performs invalid data removal operations on historical abnormal transaction data and transaction work order data according to the preset data completeness and obtains valid sample data. The valid sample data is then subjected to feature filtering operations to obtain training sample data. The second risk control submodule can determine whether the current transaction matches the user profile of the target user based on the target user tags in the account basic information data; If the current transaction does not match the user profile of the target user, then the target transaction will be handled as an exception. When the current transaction passes the transaction parameter verification, the target account information to which the current transaction belongs is obtained, and the target consumption information corresponding to the target user is determined based on the target user tag in the target account information; the current transaction information of the current transaction is compared with the target consumption information to determine whether the current transaction matches the user profile of the target user.

3. The intelligent claims settlement abnormal transaction risk control system for the insurance industry according to claim 2, characterized in that: The first risk control submodule acquires the aforementioned training sample data and constructs an abnormal transaction identification model based on logistic regression. By inputting basic account information data into the abnormal transaction identification model, it obtains the abnormal transaction identification result and outputs the first abnormal transaction identification result.

4. The intelligent claims settlement abnormal transaction risk control system for the insurance industry according to claim 2, characterized in that: The fourth risk control submodule also determines multiple transaction links associated with the transaction account to be detected when it is determined to be an abnormal transaction account, and extracts link features from the account features of the transaction account in each transaction link, and determines whether the link features are abnormal link features or normal link features.

5. The intelligent claims settlement abnormal transaction risk control system for the insurance industry according to claim 1, characterized in that: The risk control subsystem also includes a risk control evaluation submodule. The risk control evaluation submodule performs index evaluation based on the first abnormal transaction identification result, the abnormal transaction account determination result, and the user profile determination result. When the evaluation index is greater than or equal to the preset alarm threshold, a corresponding alarm prompt message is generated.

Images