A gang fraud identification method and system based on an association graph

Abstract

The present application relates to the technical field of gang fraud identification, in particular to a gang fraud identification method and system based on an association graph. The system comprises a processing and constructing unit, an analysis and identification unit, a risk assessment unit, and a risk decision unit. The analysis and identification unit identifies gang fraud behavior according to a risk strategy, transmits the identified gang fraud behavior data into the risk assessment unit, establishes a risk assessment model according to the obtained various data and the identified gang fraud behavior data, and then assesses the gang fraud risk according to the risk assessment model. The risk decision unit formulates a risk strategy according to the constructed association graph data and the assessed gang fraud risk data, reduces the risk of a single data source, and then transmits the risk strategy data into the analysis and identification unit to identify gang fraud behavior. This can effectively detect and identify gang fraud behavior that is deeply hidden, thereby improving the detection rate of fraud behavior.

Description

Technical Field

[0001] This invention relates to the field of gang fraud identification technology, and more specifically, to a gang fraud identification method and system based on association graphs. Background Technology

[0002] In real-world scenarios, as sales of new energy vehicles increase year by year, the accident rate of new energy vehicle insurance is far higher than that of fuel vehicles, and the average claim amount for new energy vehicles is also far higher than that for fuel vehicles, causing huge losses to car insurance underwriters and customers. When fraud gangs commit fraud, due to the relatively singular data source, it is difficult to effectively identify deeply hidden fraud gangs. Moreover, this is a conventional identification method that has been well known to fraud gangs, leading to more and more gangs being able to circumvent the rules of relying on historical accident data to build correlations through more means. In order to effectively identify gang fraud, we provide a gang fraud identification method and system based on correlation graphs. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for identifying gang fraud based on association graphs, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a method and system for identifying gang fraud based on association graphs, comprising the following method steps:

[0005] S1. The processing and construction unit acquires and processes vehicle network big data, historical claims data of accident-related parties, and various relevant data from vehicle insurance underwriting institutions, and then constructs a correlation graph based on the preprocessed data.

[0006] S2. The analysis and identification unit analyzes abnormal behavior based on the various data obtained from the processing and construction unit and the constructed correlation graph data. Then, based on the correlation graph data constructed in the processing and construction unit and the analyzed abnormal behavior data, it identifies gang fraud behavior. The identified gang fraud behavior data is then transmitted to the risk assessment unit, and the gang fraud risk is assessed based on the various data obtained from the processing and construction unit and the gang fraud behavior data identified in the analysis and identification unit.

[0007] S3. The risk decision-making unit formulates risk strategies based on the correlation graph data constructed in the processing and construction unit and the gang fraud risk data assessed in the risk assessment unit. The risk strategy data is then transmitted to the analysis and identification unit for identification of gang fraud behavior. The re-identified gang fraud behavior data is then transmitted to the risk assessment unit, which assesses the risk of gang fraud behavior based on the re-identified gang fraud behavior data and various other acquired data.

[0008] As a further improvement to this technical solution, it includes a processing and construction unit, an analysis and identification unit, a risk assessment unit, and a risk decision-making unit;

[0009] The processing and construction unit is used to acquire vehicle network big data and historical claims data of accident-related parties in real time, as well as various relevant data of vehicle insurance underwriting institutions, and to process and construct correlation graphs of the multiple data acquired in real time.

[0010] The analysis and identification unit is used to receive and process various data obtained from the construction unit and the constructed association graph data to analyze abnormal behavior and identify gang fraud behavior.

[0011] The risk assessment unit is used to receive various data obtained from the processing and construction unit and the data on gang fraud identified by the analysis and identification unit, and to establish a risk assessment model, and then to assess the risk of gang fraud based on the risk assessment model.

[0012] The risk decision-making unit is used to receive and process the association graph data constructed in the construction unit and the gang fraud risk data assessed in the risk assessment unit, and to formulate risk strategies.

[0013] The risk decision-making unit formulates risk strategies based on the constructed correlation graph data and the assessed gang fraud risk data. The risk strategy data is then transmitted to the analysis and identification unit, which identifies gang fraud behavior based on the risk strategy data. The identified gang fraud behavior data is then transmitted to the risk assessment unit. The risk assessment unit establishes a risk assessment model based on the acquired data and the identified gang fraud behavior data, and then assesses the gang fraud risk based on the risk assessment model.

[0014] As a further improvement to this technical solution, the processing construction unit includes an acquisition processing module and a map construction module;

[0015] The acquisition and processing module utilizes multiple channels to acquire real-time vehicle network big data and historical claims data of accident-related parties, as well as various relevant data from vehicle insurance underwriting institutions. It stores the acquired data in a database and processes the data from the database.

[0016] The graph construction module is used to receive various types of data processed in the acquisition and processing module, and determine the nodes, establish the edges and determine the weights based on the processed data, and then construct the association graph based on the node data, edge data and weight data.

[0017] When the number of multiple data points acquired in real time exceeds the set data number, it indicates that new data has been acquired. The database is then updated based on the acquired new data, and the associated map data is updated in real time.

[0018] As a further improvement to this technical solution, the analysis and identification unit includes a data analysis module and a gang identification module;

[0019] The data analysis module is used to receive various types of data acquired by the acquisition and processing module and the association graph data constructed by the graph construction module, and to analyze abnormal behavior based on the acquired various types of data and the constructed association graph data.

[0020] The gang identification module is used to receive the association graph data constructed in the graph construction module and the abnormal behavior data analyzed in the data analysis module. It uses a label propagation algorithm to identify fraudulent behavior based on the constructed association graph data, and then uses a strong connected subgraph algorithm to determine the existence of gangs based on the connection relationships between nodes in the constructed association graph data, the edge weight factors, and the analyzed abnormal behavior data. Finally, it identifies gang fraudulent behavior based on the identified fraudulent behavior and the determination of the existence of gangs.

[0021] As a further improvement to this technical solution, the risk assessment unit is used to receive various data acquired by the acquisition and processing module and gang fraud behavior data identified by the gang identification module, and to establish a risk assessment model based on the acquired various data and the identified gang fraud behavior data, and then use a fraud assessment algorithm to assess the gang fraud risk based on the risk assessment model data.

[0022] As a further improvement to this technical solution, the risk decision unit is used to receive the association graph data constructed in the graph construction module and the gang fraud risk data assessed in the risk assessment unit, and to formulate risk strategies based on the constructed association graph data and the assessed gang fraud risk data using a risk strategy algorithm.

[0023] As a further improvement to this technical solution, the tag propagation algorithm in the gang identification module is implemented as follows:

[0024] Principle 1: First, define a unique attribute label for each node in the association graph, and then spread the association through these attribute labels during the iteration process.

[0025] Principle 2: When a node selects the label that appears most frequently among its associated nodes, since each neighboring node has the same number of labels, it will randomly select a neighboring node's label as its own.

[0026] Principle 3: Each node repeats itself until the convergence condition is met, that is, the label of each node no longer changes. At this point, nodes with the same label are grouped into a group or gang.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] 1. In this method and system for identifying gang fraud based on association graphs, the risk decision-making unit formulates risk strategies by constructing association graph data and assessing gang fraud risk data, reducing the risk of relying on a single data source. The risk strategy data is then input into the analysis and identification unit for identification of gang fraud behavior. This can effectively detect and identify those deeply hidden gang fraud behaviors, thereby improving the detection rate of fraud behavior.

[0029] 2. In this method and system for identifying gang fraud based on correlation graphs, the risk assessment unit assesses the risk of gang fraud by using identified gang fraud behavior data and various acquired data. This can reveal deeper patterns and rules of gang fraud behavior. Identifying gang fraud behavior data provides an initial clue, and by associating it with other data, more hidden gang fraud patterns can be discovered, further improving the effectiveness of risk assessment. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating the overall method steps of the present invention;

[0031] Figure 2 This is an overall block diagram of the present invention;

[0032] Figure 3 This is a block diagram of the processing unit of the present invention;

[0033] Figure 4 This is a block diagram of the analysis and identification unit of the present invention;

[0034] Figure 5 A diagram illustrating the relationship between vehicles, objects, and people;

[0035] Figure 6 This is a connected subgraph.

[0036] The meanings of the labels in the diagram are as follows:

[0037] 1. Processing building units; 11. Acquisition and processing module; 12. Atlas construction module;

[0038] 2. Analysis and identification unit; 21. Data analysis module; 22. Gang identification module;

[0039] 3. Risk assessment unit; 4. Risk decision-making unit. Detailed Implementation

[0040] The technical solutions in 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.

[0041] Example 1

[0042] Considering that fraud gangs often rely on a single data source during their operations, making it difficult to effectively identify deeply hidden fraud gangs, and that this is a common identification method known to fraud gangs, more and more gangs can circumvent the rules of relying on historical incident data to build relationships through various means, we provide a method and system for identifying gang fraud based on association graphs in order to effectively identify gang fraud.

[0043] This invention provides a method and system for identifying gang fraud based on correlation graphs. Please refer to [link / reference]. Figures 1-4 The method includes the following steps:

[0044] S1. Processing and building unit 1 acquires and processes vehicle network big data, historical claims data of accident-related parties, and various relevant data of vehicle insurance underwriting institutions, and then constructs a correlation graph based on the preprocessed data;

[0045] S2. The analysis and identification unit 2 analyzes abnormal behavior based on the various data obtained from the processing and construction unit 1 and the constructed correlation graph data. Then, it identifies gang fraud based on the correlation graph data constructed from the processing and construction unit 1 and the analyzed abnormal behavior data. The identified gang fraud data is then transmitted to the risk assessment unit 3, and the gang fraud risk is assessed based on the various data obtained from the processing and construction unit 1 and the gang fraud data identified from the analysis and identification unit 2.

[0046] S3. Risk decision-making unit 4 formulates risk strategies based on the correlation graph data constructed in processing and construction unit 1 and the gang fraud risk data assessed in risk assessment unit 3. The risk strategy data is then transmitted to analysis and identification unit 2 for identification of gang fraud behavior. The re-identified gang fraud behavior data is then transmitted to risk assessment unit 3. Risk assessment unit 3 assesses the risk of re-gang fraud based on the re-identified gang fraud behavior data and various acquired data.

[0047] Processing unit 1, analysis and identification unit 2, risk assessment unit 3, risk decision-making unit 4;

[0048] The processing unit 1 is used to acquire real-time big data of the Internet of Vehicles and historical claims data of accident-related parties, as well as various relevant data of the car insurance underwriting institution, and to process and construct the correlation graph of the acquired data. The analysis and identification unit 2 is used to receive the various data acquired by the processing unit 1 and the correlation graph data constructed to analyze abnormal behavior and identify gang fraud behavior. The risk assessment unit 3 is used to receive the various data acquired by the processing unit 1 and the gang fraud behavior data identified by the analysis and identification unit 2, and to establish a risk assessment model. Then, the gang fraud risk is assessed based on the risk assessment model. The risk decision unit 4 is used to receive the correlation graph data constructed by the processing unit 1 and the gang fraud risk data assessed by the risk assessment unit 3, and to formulate risk strategies.

[0049] Risk decision-making unit 4 formulates risk strategies based on the constructed correlation graph data and the assessed gang fraud risk data. The risk strategy data is then transmitted to analysis and identification unit 2, which identifies gang fraud behavior based on the risk strategy data. The identified gang fraud behavior data is then transmitted to risk assessment unit 3. Risk assessment unit 3 establishes a risk assessment model based on the acquired data and the identified gang fraud behavior data, and then assesses the gang fraud risk based on the risk assessment model. This provides scientific and objective assessment results. Compared with subjective judgment or simple rules, the risk assessment model can more comprehensively and accurately consider multiple factors, thereby more accurately assessing the probability of gang fraud risk.

[0050] The following is a more detailed explanation of the above units; please refer to [link / reference]. Figures 1-4 ;

[0051] Processing unit 1 includes an acquisition and processing module 11 and a map construction module 12;

[0052] The acquisition and processing module 11 acquires real-time vehicle network big data and historical claims data of accident-related parties through multiple channels, as well as various relevant data of vehicle insurance underwriting institutions. It stores the acquired data in a database, retrieves and processes various data from the database, reduces data bias and improves data quality by processing the acquired data, and then transmits the processed data to the graph construction module 12.

[0053] The graph construction module 12 receives various types of data processed by the acquisition and processing module 11, and determines the nodes, establishes the edges, and determines the weights based on the processed data. Then, it constructs the association graph based on the node data, edge data, and weight data. When constructing the association graph, it is necessary to consider the association relationship and association strength between different subjects to facilitate the construction of a high-quality graph. When the number of various types of data acquired in real time is greater than the set number of data, it indicates that new data has been acquired. The database is updated based on the acquired new data, and the association graph data is updated in real time. The constructed or updated association graph data is then transmitted to the data analysis module 21.

[0054] In the graph construction module 12, the determination of nodes, the establishment of edges, and the determination of weights are crucial. Node determination refers to identifying entities in the transaction data that should be represented as nodes. Entities include individuals, organizations, and accounts. Edge establishment refers to establishing connections between nodes based on the relationships in the transaction data, representing the interactions or associations between nodes. When there are transaction records between two nodes, an edge can be established to represent the relationship between them. The determination of weights is used to represent the strength of the edges, which can be determined based on the characteristics in the transaction data. In addition to transaction data, transaction frequency, amount, and time interval can also be used. At the same time, it can also reflect the degree of association between nodes to judge the closeness of the fraud gang relationship.

[0055] For example, a vehicle-object-person relationship graph can be constructed based on offline batch data and real-time streaming data.

[0056] The offline batch data comes from insurance company claims data, and mainly includes three types of information: vehicles, goods, and people.

[0057] Real-time streaming data mainly infers the relationship between the claim vehicle and third-party vehicles based on real-time data from the Internet of Vehicles (IoV) big data. Its main contents include whether the location data of the claim vehicle and the third-party vehicle are frequently together, whether the relationship between the people involved is that of relatives or friends, and whether there is any suspected group insurance fraud.

[0058] Based on the acquired data, a relationship graph surrounding vehicles, objects, and people can be constructed, such as... Figure 5 As shown;

[0059] The general public includes the insured, the person who reports the incident, the policyholder, the driver of the insured vehicle and the third-party vehicle, the injured party in the insured vehicle and the third-party vehicle, and the payer;

[0060] The term "vehicle" includes the vehicle in question and vehicles owned by third parties.

[0061] The items include types of damage.

[0062] The essence of this relationship graph is to integrate information on claims cases provided by insurance companies. In anti-fraud scenarios, in addition to considering the attributes of individual information points, the hidden connections between customers often contain more valuable information. Through the relationship graph, we can understand the relationship between vehicles and people and the relationship between people in claims cases. The information of people can be obtained from the information provided by the insurance company to see if there is a kinship relationship between people. The biggest advantage of using the relationship graph for anti-fraud identification is that it can identify groups of people who are normal individuals but are fraudulent. Individual relationships include kinship relationships. Further, we can identify the relationship between two people by the consistency of their registered family information or by clarifying that they are family members or colleagues of the applicant based on the application information. We can also associate them by the names of their past companies. If the names are the same, there is a risk of fraud. We can also associate them by mobile phone numbers, including sharing mobile phone numbers and setting emergency contacts, which both indicate a risk of fraud. Finally, we can identify them by financial transaction relationships. People who have transferred money through banks or third-party apps are at risk.

[0063] Analysis and identification unit 2 includes a data analysis module 21 and a gang identification module 22;

[0064] The data analysis module 21 receives various data acquired by the acquisition and processing module 11 and the association graph data constructed by the graph construction module 12. Based on the acquired data and the constructed association graph data, it analyzes abnormal behavior. Through clustering of multiple data sources and construction of association graphs, it can discover patterns and relationships inconsistent with normal behavior, and can identify improper behaviors such as fraudulent transactions and fraudulent transactions. The analyzed abnormal behavior data is then transmitted to the gang identification module 22 for specific analysis.

[0065] Scenario 1: Based on user behavior, transaction behavior, and identity verification data, fake identities and fraudulent activities can be identified;

[0066] Scenario 2: Based on the behavioral data of employees or partners, internal fraud and data breaches can be detected and prevented;

[0067] Situation ③: Transaction behavior and fund flow data can identify money laundering and fund transfer activities.

[0068] The gang identification module 22 is used to receive the association graph data constructed in the graph construction module 12 and the abnormal behavior data analyzed in the data analysis module 21. It uses the label propagation algorithm to identify fraudulent behavior based on the constructed association graph data, and then uses the strong connected subgraph algorithm to determine the existence of gangs based on the connection relationship between nodes in the constructed association graph data, the edge weight factors, and the analyzed abnormal behavior data. Finally, it identifies gang fraudulent behavior based on the identified fraudulent behavior and the determination that gangs exist.

[0069] The implementation principle of the tag propagation algorithm in the gang identification module 22:

[0070] Principle 1: First, define a unique attribute label for each node in the association graph, and then spread the association through these attribute labels during the iteration process.

[0071] Principle 2: When a node selects the label that appears most frequently among its associated nodes, since each neighboring node's label appears the same number of times, it will randomly select a neighboring node's label as its own label.

[0072] Principle 3: Each node repeats itself until the convergence condition is met, that is, the label of each node no longer changes. At this point, nodes with the same label are grouped into a group or gang.

[0073] Implementation process of the strongly connected subgraph algorithm:

[0074] Process ①: If there is at least one path between two vertices, the two vertices are said to be strongly connected.

[0075] Step 2: Further, if for a subgraph in the network, there is at least one path between every two vertices, then this subgraph is considered a strongly connected subgraph.

[0076] Process ③, such as Figure 6 Nodes 1, 2, 3, and 4 of the connected subgraph form a strongly connected subgraph, but nodes 5 and 6 are not included in this strongly connected subgraph.

[0077] The risk assessment unit 3 is used to receive various data acquired by the acquisition and processing module 11 and the gang fraud behavior data identified by the gang identification module 22, and to establish a risk assessment model based on the acquired data and the identified gang fraud behavior data. The collection of various data and the identified gang fraud behavior data can serve as the basis for model training. These data may include the characteristics of fraud behavior, historical fraud cases, and the behavior patterns of fraudsters. Through these data, a more accurate risk assessment model can be established, and then the fraud assessment algorithm can be used to assess the gang fraud risk based on the risk assessment model data.

[0078] The steps to establish a risk assessment model are as follows:

[0079] ① First, collect and obtain various types of data and identify group fraud behavior data, and initialize cluster values;

[0080] ② Then, assign the data values ​​of the fraudulent activities of the groups that are not clustered to the cluster set that is closest to them;

[0081] ③ Repeatedly cluster the current data values ​​of gang fraud behavior obtained through clustering and the multiple data values ​​obtained from the initial clustering until the final standard is reached, and then stop to obtain the risk assessment model.

[0082] Fraud assessment algorithm formula:

[0083] Sp = (Tp + Tn) / (Tp + Fp + Tn + Fn);

[0084] Where Sp represents the accuracy rate of the assessed gang fraud risk, Tp refers to the number of true positives, that is, the number of samples in which the risk assessment model correctly predicts a positive example as a positive example, Tn refers to the number of true positives, that is, the number of samples in which the model correctly predicts a negative example as a negative example, Fp refers to the number of false positives, that is, the number of samples in which the model incorrectly predicts a negative example as a positive example, and Fn refers to the number of false negatives, that is, the number of samples in which the model incorrectly predicts a positive example as a negative example. This formula is used to assess gang fraud risk. By assessing gang fraud risk through the risk assessment model, potential risks can be predicted more accurately, the accuracy of early warning can be effectively improved, and fraudulent activities can be prevented and combated.

[0085] Then, the accuracy rate of the set group fraud risk and the accuracy rate of the assessed group fraud risk are used to make a judgment on the specific situation;

[0086] Scenario 1: When the accuracy rate of the assessed group fraud risk is greater than the set accuracy rate of the group fraud risk, it indicates that the group fraud risk level is high.

[0087] Scenario 2: When the accuracy rate of the assessed gang fraud risk is less than the set accuracy rate of the gang fraud risk, it indicates that the gang fraud risk level is low.

[0088] The risk decision-making unit 4 is used to receive the association graph data constructed in the graph construction module 12 and the gang fraud risk data assessed in the risk assessment unit 3, and to formulate risk strategies based on the constructed association graph data and the assessed gang fraud risk data using the risk strategy algorithm.

[0089] Risk strategy algorithm formula:

[0090] F(x)=Δχ×y×Δγ×(x1+x2+...+x n );

[0091] Where F(x) represents the specified risk strategy, Δχ refers to the original data value of the constructed association graph, y refers to the updated association graph data value, Δγ refers to the accuracy of the assessed gang fraud risk, and x1+x2+...+x n This refers to the different variable values ​​generated during the assessment of gang fraud risk. This formula is used to formulate risk strategies. By analyzing correlation graph data and gang fraud risk data, potential risk chains in fraud networks can be discovered. In fraud networks, there may be complex connections and interactions between different individuals. By analyzing these connections and transaction patterns, potential risk chains can be identified, helping to formulate targeted risk strategies.

[0092] Usage process:

[0093] The gang identification module 22 identifies gang fraud based on the constructed association graph data and analyzed abnormal behavior data. It then transmits the identified gang fraud to the risk assessment unit 3 and establishes a risk assessment model and assesses the gang fraud risk based on the acquired data and the identified gang fraud data. The assessed gang fraud risk data is then transmitted to the risk decision-making unit 4, which formulates risk strategies based on the constructed association graph data and the assessed gang fraud risk data. The risk strategy data is then transmitted to the gang identification module 22, which identifies gang fraud again based on the risk strategy data. Finally, the re-identified gang fraud data is transmitted to the risk assessment unit 3 for a re-assessment of gang fraud risk.

[0094] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method and system for identifying gang fraud based on association graphs, characterized in that: The methods and steps include the following: S1. Processing and construction unit (1) acquires and processes the Internet of Vehicles big data, historical claims data of accident-related parties and various relevant data of car insurance underwriting institutions, and then constructs the association graph based on the preprocessed data; S2. The analysis and identification unit (2) analyzes abnormal behavior based on the various data obtained in the processing and construction unit (1) and the constructed association graph data. Then, it identifies gang fraud behavior based on the association graph data constructed in the processing and construction unit (1) and the analyzed abnormal behavior data. The identified gang fraud behavior data is then transmitted to the risk assessment unit (3) and the gang fraud risk is assessed based on the various data obtained in the processing and construction unit (1) and the gang fraud behavior data identified in the analysis and identification unit (2). S3. The risk decision-making unit (4) formulates risk strategies based on the correlation graph data constructed in the processing and construction unit (1) and the gang fraud risk data assessed in the risk assessment unit (3). The risk strategy data is then transmitted to the analysis and identification unit (2) for identification of gang fraud behavior. The re-identified gang fraud behavior data is then transmitted to the risk assessment unit (3). The risk assessment unit (3) assesses the risk of gang fraud behavior based on the re-identified gang fraud behavior data and the various data obtained.

2. The method and system for identifying gang fraud based on association graphs according to claim 1, characterized in that: It includes a processing and construction unit (1), an analysis and identification unit (2), a risk assessment unit (3), and a risk decision-making unit (4); The processing construction unit (1) is used to acquire vehicle network big data and historical claims data of accident-related parties in real time, as well as various relevant data of vehicle insurance underwriting institutions, and to process and construct association graphs of the multiple data acquired in real time. The analysis and identification unit (2) is used to receive various data and constructed association graph data obtained by the processing and construction unit (1) to analyze abnormal behavior and identify gang fraud behavior; The risk assessment unit (3) is used to receive various data obtained from the processing construction unit (1) and the gang fraud behavior data identified by the analysis and identification unit (2) and establish a risk assessment model, and then assess the gang fraud risk based on the risk assessment model. The risk decision-making unit (4) is used to receive the association graph data constructed in the processing construction unit (1) and the gang fraud risk data assessed in the risk assessment unit (3) and to formulate risk strategies. The risk decision-making unit (4) formulates risk strategies based on the constructed association map data and the assessed gang fraud risk data, transmits the risk strategy data to the analysis and identification unit (2) and identifies gang fraud behavior based on the risk strategy data, and then transmits the identified gang fraud behavior data to the risk assessment unit (3). The risk assessment unit (3) establishes a risk assessment model based on the acquired data and the identified gang fraud behavior data, and then assesses the gang fraud risk based on the risk assessment model.

3. The method and system for identifying gang fraud based on association graphs according to claim 2, characterized in that: The processing construction unit (1) includes an acquisition processing module (11) and a map construction module (12); The acquisition and processing module (11) uses multiple channels to acquire real-time big data of the Internet of Vehicles and historical claims data of accident-related parties, as well as various relevant data of the car insurance underwriting institution, stores the acquired data in the database, and processes the data from the database. The graph construction module (12) is used to receive various types of data processed in the acquisition and processing module (11), and to determine the nodes, establish the edges and determine the weights based on the various types of data processed, and then construct the association graph based on the node data, edge data and weight data. When the number of multiple data points acquired in real time exceeds the set data number, it indicates that new data has been acquired. The database is then updated based on the acquired new data, and the associated map data is updated in real time.

4. The method and system for identifying gang fraud based on association graphs according to claim 3, characterized in that: The analysis and identification unit (2) includes a data analysis module (21) and a gang identification module (22); The data analysis module (21) is used to receive various data acquired in the acquisition and processing module (11) and the association map data constructed in the map construction module (12), and to analyze abnormal behavior based on the acquired various data and the constructed association map data; The gang identification module (22) is used to receive the association graph data constructed in the graph construction module (12) and the abnormal behavior data analyzed in the data analysis module (21), and use the label propagation algorithm to identify the existence of fraudulent behavior based on the constructed association graph data. Then, the strong connected subgraph algorithm is used to determine the existence of gangs based on the connection relationship between nodes in the constructed association graph data, the weight factor of the edges, and the analyzed abnormal behavior data. Finally, the gang fraudulent behavior is identified based on the identified fraudulent behavior and the determination of the existence of gangs.

5. The method and system for identifying gang fraud based on association graphs according to claim 4, characterized in that: The risk assessment unit (3) is used to receive various data acquired in the acquisition and processing module (11) and gang fraud behavior data identified in the gang identification module (22), and to establish a risk assessment model based on the acquired various data and the identified gang fraud behavior data, and then use a fraud assessment algorithm to assess the gang fraud risk based on the risk assessment model data.

6. The method and system for identifying gang fraud based on association graphs according to claim 5, characterized in that: The risk decision unit (4) is used to receive the association graph data constructed in the graph construction module (12) and the gang fraud risk data assessed in the risk assessment unit (3), and to formulate risk strategies based on the constructed association graph data and the assessed gang fraud risk data using the risk strategy algorithm.

7. The method and system for identifying gang fraud based on association graphs according to claim 6, characterized in that: The tag propagation algorithm in the gang identification module (22) is implemented as follows: Principle 1: First, define a unique attribute label for each node in the association graph, and then spread the association through these attribute labels during the iteration process. Principle 2: When a node selects the label that appears most frequently among its associated nodes, since each neighboring node's label appears the same number of times, it will randomly select a neighboring node's label as its own label. Principle 3: Each node repeats itself until the convergence condition is met, that is, the label of each node no longer changes. At this point, nodes with the same label are grouped into a group or gang.

Images