Social network-based insurance fraud ring identification method, system, device and medium

Abstract

The application discloses a social network-based insurance fraud gang identification method, system, device and medium, and relates to the technical field of computers, wherein the method comprises the following steps: constructing a social network relationship graph based on the role information and subject information in a newly added claim event and using insurance business relationships; obtaining traversal starting points for constructing a closed-loop relationship graph according to the social network relationship graph; constructing a plurality of closed-loop relationship graphs for a plurality of traversal starting points by using a first adjacent point acquisition method based on claim event relationships; and obtaining a target closed-loop relationship graph by merging the plurality of closed-loop relationship graphs by using a second adjacent point acquisition method based on claim event relationships, wherein the node information in the target closed-loop relationship graph is used to represent a target insurance fraud gang. The application is suitable for efficient investigation of fraud case identification and improves the accuracy and efficiency of fraud case identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular to an insurance fraud gang identification method, system, device and medium based on a social network. BACKGROUND

[0002] In recent years, the insurance industry has continuously played the role of economic shock absorber and social stabilizer, and has served and protected people's lives. With the innovation and development of insurance products and services, insurance fraud methods have also been upgraded. The current insurance fraud risk is showing a trend of concealment, organization and gang. Therefore, effectively identifying insurance fraud gangs has become an important proposition for current insurance anti-fraud.

[0003] At present, domestic fraud gang identification using data mining mainly falls into two categories: one is a regular rule method, and the other is a machine learning method (fraud identification model is generated by training samples of fraud and non-fraud). Both methods have certain limitations. The regular rule method refers to analyzing existing fraud cases, manually extracting key features, and subsequently identifying new cases that trigger existing case features. The disadvantages of this method are that it is not timely and cannot identify new fraud cases, and criminals can also learn to avoid key features, making the effectiveness of the rules short-lived. The machine learning method refers to using algorithms such as CNN and LSTM to train models by labeling fraud and non-fraud cases to generate fraud identification models for detecting and identifying new cases. The disadvantage of this method is that the sample size of fraud cases is small, resulting in an inappropriate ratio of fraud and non-fraud samples in the training samples, which leads to insufficient accuracy of the model in identifying fraud cases. SUMMARY

[0004] Therefore, the present application provides an insurance fraud gang identification method, system, device and medium based on a social network, which mainly aims to solve the technical problems that the existing regular rule method can only identify fraud cases with key features, and the scheme has low timeliness, and the machine learning method has high requirements for sample data, resulting in low accuracy of fraud case identification.

[0005] According to one aspect of the present application, an insurance fraud gang identification method based on a social network is provided, which comprises:

[0006] Based on the role information and subject information in the new claim settlement event, a social network relationship graph is constructed using insurance business relationships;

[0007] According to the social network relationship graph, an traversal starting point for constructing a closed-loop relationship graph is obtained;

[0008] The first neighbor point acquisition method based on the claim event relationship is used to construct a plurality of closed loop relationship graphs for the plurality of traversal starting points.

[0009] The second neighbor point acquisition method based on the claim event relationship is used to obtain a target closed loop relationship graph by merging the plurality of closed loop relationship graphs, and the node information in the target closed loop relationship graph is used to represent a target insurance fraud gang.

[0010] According to another aspect of the present application, a social network-based insurance fraud gang identification system is provided, which comprises:

[0011] The first construction module is configured to construct a social network relationship graph based on the role information and the subject information in the new claim event and the insurance business relationship;

[0012] The traversal module is configured to obtain traversal starting points for constructing closed loop relationship graphs according to the social network relationship graph;

[0013] The second construction module is configured to construct a plurality of closed loop relationship graphs for the plurality of traversal starting points by using a first neighbor point acquisition method based on the claim event relationship;

[0014] The merging module is configured to obtain a target closed loop relationship graph by merging the plurality of closed loop relationship graphs by using a second neighbor point acquisition method based on the claim event relationship, and the node information in the target closed loop relationship graph is used to represent a target insurance fraud gang.

[0015] According to still another aspect of the present application, a computer storage medium is provided, which stores a computer program, and the program is executed by a processor to implement the above-mentioned social network-based insurance fraud gang identification method.

[0016] According to yet another aspect of the present application, a computer device is provided, which comprises a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, and the processor executes the program to implement the above-mentioned social network-based insurance fraud gang identification method.

[0017] By the technical scheme, the social network-based insurance fraud gang identification method, system, device and medium provided by the application can identify fraud cases with key features, and the scheme has low timeliness. Compared with the technical scheme in which the fraud case identification accuracy is low due to high requirements for sample data based on the machine learning method, the application constructs a social network relationship graph based on the role information and subject information in the new claim settlement event, and uses the insurance business relationship. According to the social network relationship graph, the traversal starting point for constructing the closed loop relationship graph is obtained. The first neighbor point acquisition method based on the claim settlement event relationship is used to construct a plurality of closed loop relationship graphs for a plurality of traversal starting points. The second neighbor point acquisition method based on the claim settlement event relationship is used to obtain the target closed loop relationship graph by merging the plurality of closed loop relationship graphs. The node information in the target closed loop relationship graph is used to represent the target insurance fraud gang. As can be seen, the social network relationship graph constructed based on the role information and subject information in the new claim settlement event is analyzed by traversing each node in the social network relationship graph, and the correlation between each node is obtained based on the claim settlement event relationship, so as to obtain the target closed loop relationship graph. That is, the target insurance fraud gang information is found through the target closed loop relationship graph, which can not only avoid the limitation of the historical key features in the regular rule method, but also ensure the identification accuracy of the fraud case.

[0018] The above description is only a summary of the technical scheme of the application. In order to more clearly understand the technical means of the application, the specific embodiments of the application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0019] The drawings described herein are used to provide further understanding of the application, and form a part of the application. The schematic embodiments of the application and their descriptions are used to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0020] Figure 1 A flowchart of a social network-based insurance fraud gang identification method provided by an embodiment of the application is shown;

[0021] Figure 2 A flowchart of another social network-based insurance fraud gang identification method provided by an embodiment of the application is shown;

[0022] Figure 3 A schematic diagram of a target closed loop relationship graph of another social network-based insurance fraud gang identification method provided by an embodiment of the application is shown;

[0023] Figure 4A structural schematic diagram of an insurance fraud gang identification system based on a social network is shown.

[0024] Figure 5 A structural schematic diagram of another insurance fraud gang identification system based on a social network is shown. DETAILED DESCRIPTION

[0025] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0026] The existing regular rule-based method can only identify fraud cases with key features, and the timeliness of the scheme is low. The machine learning-based method has a higher requirement for sample data, and thus has the technical problem of low accuracy in identifying fraud cases. The present embodiment provides an insurance fraud gang identification method based on a social network, which can analyze the association relationship in each claim event relationship based on a new claim event, so as to obtain a target closed-loop relationship graph, that is, to find target insurance fraud gang information through the target closed-loop relationship graph, which not only avoids the limitation of historical key features in the regular rule method, but also ensures the accuracy of fraud case identification. As shown in Figure 1 The above method includes the following steps:

[0027] Step 101, based on the role information and the subject information in the new claim event, a social network relationship graph is constructed using the insurance business relationship.

[0028] In the present embodiment, the traditional social network is a collection of social actors and their relationships. According to the needs of the actual application scene, the social network in the present embodiment is a collection of social actors with insurance business relationships and their related insurance subject information. Specifically, nodes are created based on the role information and the vehicle information in the subject information, and the vehicle information in the role information and the subject information in the new claim event is used as the traversal starting point for constructing the social network relationship graph. By associating other new claim events and the nodes in the new claim events and historical claim events, a social network relationship graph for the new claim event is obtained, which not only enables further determination of historical suspected fraud cases, but also enables mining of fraud cases in new claim events.

[0029] The different role information in the new claim event can include a driver, an owner, a third party (an insured, an insured person, a claimant, a beneficiary, a claimant, an insured person, a claimant, an investigator, etc.), and the target information can include a vehicle, a claim amount of the claim event, etc. The insurance business relationship can include an insured-insured relationship (a policy relationship), an interest-related person relationship in the claim event (an insured person-claimant relationship, a beneficiary-insured person relationship), other related person relationships in the claim event (a claimant, an investigator), and other relationships related to the insurance business. According to the requirements of the actual application scene, the role information can further include enterprise information associated with the personnel, that is, the unified social credit code of the personnel-associated enterprise as a unique identifier.

[0030] Step 102, obtaining a traversal starting point for constructing a closed-loop relationship graph according to the social network relationship graph.

[0031] In the embodiment, in order to ensure the fullness of the constructed closed-loop relationship graph, each node in the social network relationship graph is traversed, and a node corresponding to a maximum value of the number of claim events is taken as a traversal starting point for constructing a closed-loop relationship graph; or a node corresponding to a maximum value of the number of claim events and a maximum value of claim amount information is taken as a traversal starting point for constructing a closed-loop relationship graph. The maximum value of the claim amount information refers to a sum of compensation amounts of each claim event in the node, and the sum of the compensation amounts is the maximum value in each node of the social network relationship graph.

[0032] Step 103, constructing a plurality of closed-loop relationship graphs for a plurality of traversal starting points by using a first neighbor point acquisition mode based on a claim event relationship.

[0033] In the embodiment, after the traversal starting point is determined, other nodes are traversed, a node having the same insurance business relationship with the traversal starting point is obtained and taken as a first neighbor point of the traversal starting point, and after the number of claim events stored in the node and the sum of the compensation amounts of each claim event are determined, other nodes having the same insurance business relationship with the traversal starting point are continuously traversed. If the number of claim events stored in the node is greater than the first neighbor point, or the number of claim events is the same but the sum of the compensation amounts of each claim event is greater than the first neighbor point, the first neighbor point is updated, until the traversal ends, a target first neighbor point is determined, and the same is repeated to obtain a relationship chain, and then a plurality of closed-loop relationship graphs of the current traversal starting point are constructed based on the nodes in the relationship chain.

[0034] Step 104, obtaining a target closed-loop relationship graph by merging the plurality of closed-loop relationship graphs by using a second neighbor point acquisition mode based on a claim event relationship.

[0035] In the embodiment, for the multiple closed loop relationship graphs constructed for different traversal starting points, it is analyzed whether there is an associated node of the same claim event relationship in each closed loop relationship graph, and at least two closed loop relationship graphs are merged based on the associated node in the at least two closed loop relationship graphs to obtain a target closed loop relationship graph. Wherein, the role information stored in each node in the target closed loop relationship graph is used to represent a target insurance fraud gang.

[0036] For the embodiment, the above scheme can be followed to construct a social network relationship graph based on the role information and subject information in the new claim event, using the insurance business relationship; according to the social network relationship graph, the traversal starting point for constructing the closed loop relationship graph is obtained; using the first neighbor point acquisition method based on the claim event relationship, multiple closed loop relationship graphs for multiple traversal starting points are constructed; using the second neighbor point acquisition method based on the claim event relationship, a target closed loop relationship graph is obtained by merging the multiple closed loop relationship graphs, and the node information in the target closed loop relationship graph is used to represent a target insurance fraud gang. Compared with the existing regular rule method which can only identify fraud cases with key characteristics and has low timeliness, and the machine learning method which has high requirements for sample data and has low fraud case identification accuracy, the social network relationship graph constructed based on the role information and subject information in the new claim event in the embodiment is analyzed by traversing each node in the social network relationship graph, and the association relationship between different insurance roles and insurance subjects in the insurance claim event is found based on the claim event relationship, so as to obtain the target closed loop relationship graph. That is, the target insurance fraud gang information is found through the target closed loop relationship graph, which effectively improves the identification accuracy and timeliness of the insurance fraud gang.

[0037] Further, as a refinement and extension of the above embodiment, in order to fully describe the specific implementation process of the embodiment, another insurance fraud gang identification method based on social network is provided. The embodiment can be applied to a car insurance system. The embodiment can find the association relationship between different insurance roles and insurance subjects in the insurance claim event based on the claim event relationship, find the target insurance fraud gang information, and effectively improve the identification accuracy and timeliness of the insurance fraud gang. As shown in Figure 2 The method comprises the following steps:

[0038] Step 201, based on the role information and subject information in the new claim event, a social network relationship graph is constructed using the insurance business relationship.

[0039] In the implementation, in order to illustrate the specific implementation of step 201, as a preferred embodiment, the step 201 includes: taking the role information in the new claim event and the vehicle information in the subject information as nodes in the social network relationship graph; for each node, obtaining the number of claim events associated with the node and storing the number of claim events on the node; and connecting the nodes by using the insurance business relationship to construct the social network relationship graph.

[0040] In the implementation, all the information of the roles (which can also include the roles associated with enterprises) and the subjects having the insurance business relationship is traversed, and invalid information is cleaned. Specifically, the identity card numbers (which can also include the unified social credit codes of the enterprises associated with the personnel) of the insurance roles in all the insurance business relationships and the vehicle frame numbers of the vehicles are obtained as unique identifiers, and are saved into a set as nodes in the social network relationship graph. Each element identifier (corresponding to the node in the social network relationship graph) in the set has uniqueness. Each time a new element identifier is extracted, it is checked whether the element identifier exists in the set. If not, the new element identifier is added to the set. If yes, the element identifier is discarded, so as to ensure the uniqueness of the element identifiers in the set. Assuming that n element identifiers are extracted, the set finally formed is P={p1, p2, …, pn}. n}.

[0041] In the implementation, the social network relationship graph includes: a first social network relationship graph constructed by traversing the new claim event, the first social network relationship graph including the new claim event; or / and a second social network relationship graph constructed by traversing the historical claim event and the new claim event, the second social network relationship graph including the new claim event and the historical claim event.

[0042] The embodiment is different from the prior art. Without relying on specific restrictions and without needing to find samples in advance for training, the embodiment is based on the characteristics of the case of the gang engaging in insurance fraud, that is, the number of times of the obvious correlation between the roles and the vehicles is obviously higher than that in the normal case, and the probability of occurrence obviously deviates from the natural law. By periodically traversing to find the correlation between all the insurance roles and the subjects, the traversal includes the traversal of the role information and the subject information in the new claim event, so as to realize the identification of the new fraud case. The role information and the subject information in the new claim event are used to realize the traversal of the new claim event and the historical claim event, so as to realize the further determination of the historical suspected fraud case, and realize the fullness of the fraud case identification.

[0043] Step 202: obtaining a traversal starting point for constructing a closed-loop relationship graph according to the social network relationship graph.

[0044] In implementation, to illustrate the specific implementation of step 202, as a preferred embodiment, the step 202 comprises: acquiring the node corresponding to the maximum number of claim events in the social network relationship graph, and taking it as the traversal starting point for constructing the closed-loop relationship graph; or, the target information comprises the claim amount corresponding to each claim event, acquiring a plurality of nodes corresponding to the maximum number of claim events in the social network relationship graph, and determining the traversal starting point for constructing the closed-loop relationship graph according to the claim amount information in the plurality of nodes.

[0045] In implementation, the graph of the social network is called a social network relationship graph and a social network relationship matrix, which is composed of a plurality of nodes (social actors with insurance relationship) and the connection (social actors with insurance interest relationship) between nodes. The nodes in the social network relationship graph are used to represent various roles (social actors) and targets with insurance business relationship, and the edges are used to represent various insurance business relationships between roles, between roles and targets, such as the relationship between the policyholder and the insured, the relationship between the driver of the main vehicle and the third party in the accident, etc., and the relationship is undirected. Among them, the social network relationship graph is composed of a set of nodes N = {n1, n2, …, n k} and the connection L = {l1, l2, …, l m} between nodes.

[0046] According to the requirements of the actual application scene, from a certain element identifier, if the element identifier p n is involved in a claim event, the number is recorded as 1, the number of involved claim events is n, and the claim event number set is formed after traversing each element identifier. n is the number of claim events involved in the element identifier p n , and p1 = max(n) is recorded, that is, the element identifier with the largest number of claim events is taken as the traversal starting point P1. Further, if there are element identifiers with the same number of claim events, the element identifier with the largest claim amount is taken as P1, and if both conditions are the same, one element identifier is randomly selected as the starting point P1.

[0047] Step 203, constructing a relationship chain based on the current traversal starting point by using a first neighbor point acquisition method based on claim event relationship.

[0048] In implementation, the first neighbor point acquisition method is to acquire a plurality of nodes having the same insurance business relationship with the traversal starting point, and to determine the direct neighbor point of the traversal starting point according to the number of claim events corresponding to the same insurance business relationship; the direct neighbor point is that there is no other node between the traversal starting point and the neighbor point, and the indirect neighbor point is that there is another node between the traversal starting point and the neighbor point.

[0049] In practice, starting from the traversal starting point, let's say p1, we traverse the element identifiers p that have a claim event relationship with it. i The relationship between the two claims events is denoted as l 1,i , continue with p i Begin by iterating through the element identifier p that has a claim event relationship with it. i+1 At this time, p1 and p i+1 The relationship is l 1,i,i+1 Continue the above search process until the relationship no longer changes, thus obtaining a relation chain L with a unique starting point and different relation lengths. A complete relation chain is denoted as l. 1,2,…,n It should be noted that if p1 and p2 have a claim event relationship, it is denoted as l. 12 The relationship between claims events is not directional, that is... 12 =l 21 For ease of narration, a naming format is used with subscript numbers first and main numbers second.

[0050] Specifically, using the element identifier p i For example, select the element identifier with the most claims, or select the element identifier with the highest claim amount from among the element identifiers with the same number of claims. If both of these conditions are the same, then randomly select one element identifier from those that meet the conditions as p. i The relationship between the two claims events is denoted as l 1,i Similarly, continuing with the element identifier p... i Starting from this point, the above process is repeated, with already associated element identifiers no longer participating in the traversal, until no new relationship chains appear. Further, based on the above traversal order, the relationship chains are sorted to obtain a complete set of relationship chains, denoted as l. 1,2,…,n This includes the first traversal of the chain, denoted as The chain traversed in the nth iteration is denoted as l. (1,i,i+1)n All nodes in the relation chain set correspond to the nodes in the traversal chain. Based on the needs of the actual application scenario, it can be determined that the element identifiers on the first traversal chain have the greatest fraud risk. The first traversal chain is taken as the target relation chain, or the traversal chain with the most element identifiers is taken as the target relation chain. No specific limitation is made on the target relation chain here.

[0051] Step 204: Obtain the indirect neighboring nodes in the relationship chain that have the same insurance business relationship with the traversal starting point, and construct multiple closed-loop relationship graphs for the current traversal starting point based on the indirect neighboring nodes.

[0052] Step 205, using a second neighbor point acquisition method based on the claim event relationship, obtaining a target closed loop relationship graph by merging the plurality of closed loop relationship graphs.

[0053] In implementation, the node information in the target closed loop relationship graph is used to represent a target insurance fraud gang, if there are associated nodes with the same claim event relationship in at least two closed loop relationship graphs in the plurality of closed loop relationship graphs, then the at least two closed loop relationship graphs are merged based on the associated nodes to obtain a target closed loop relationship graph, the associated nodes refer to nodes with the same claim event relationship that are not in the same closed loop relationship graph.

[0054] In implementation, in a social network relationship graph, if any node has a path connected thereto, the social network relationship graph is a closed loop graph relationship, that is, by acquiring a plurality of nodes with the same insurance business relationship as the traversal starting point, and taking the node corresponding to the maximum number of claim events as the first neighbor point of the traversal starting point, and continuing to acquire the second neighbor point of the first neighbor point, until a plurality of closed loop relationship graphs for the traversal starting point are obtained. Therefore, in a complete relationship chain set l 1,2,…,n In implementation, if n has an insurance business relationship with 1, it is considered that a closed loop can be formed, for example, if n=3, it is a 3-element loop, and if n=4, it is a 4-element loop, for example, based on the policyholder a, through the insurance business relationship (underwriting relationship, claim relationship, etc.), the other roles are associated using the shortest path, thereby obtaining a closed loop relationship graph. According to the requirements of the actual application scene, when a larger loop contains a smaller loop, the elements in the small loop are considered as one of the elements of the large loop, and the elements on the small loop are determined to be high-risk fraud personnel / high-risk fraud enterprises / high-risk fraud vehicles / high-risk fraud insurance targets, thereby effectively reducing the acquisition time of the closed loop relationship graph.

[0055] Further, for two or more different closed loop relationship graphs, if there is an association between at least one element of each two closed loop relationship graphs, then the two or more closed loop relationship graphs are merged to obtain a target closed loop relationship graph for representing a target insurance fraud gang. Specifically, by calculating the association degree between the elements on the loop and all elements on the loop, the plurality of elements with the highest association degree are determined to have the highest fraud risk, see steps 206-208 for details, and the determination method is not specifically limited here.

[0056] In a social network relationship graph, when two nodes are connected by a straight line, the two nodes are called adjacent, and the nodes adjacent to a certain node are called neighbor points of the certain node, therefore the association degree can also be represented as the number of neighbor points of a certain node n i . i) According to the requirements of actual application scenarios, the correlation degree can also be density, that is, the ratio of the number of actually existing correlation relationships in the social network relationship graph to the maximum number of possible relationships. If the density of a social network relationship graph is 1, it means that each node in the social network relationship graph is connected to other nodes. Conversely, if the density of the social network relationship graph is 0, it means that any node in the social network relationship graph is not connected.

[0057] It should be noted that the new element set is formed based on the newly added elements and the elements not included in the closed loop. Firstly, it can repeat step 203 to form a new relationship chain. Secondly, it can repeat step 204 to find a new closed loop relationship graph. Thirdly, it can repeat step 205 to merge the new closed loop relationship graph (for example, the relationship chain 5-3-4, 0-1-2 in Figure 3 ) in the above table) with the existing target closed loop relationship graph to update the target closed loop relationship graph. Fourthly, for the chain not included in the target closed loop relationship graph, the relationship between the elements in the above chain and the elements in the existing target closed loop relationship graph is traversed. If there are more than or equal to two associated elements (for example, the relationship chain 5-3-4 in Figure 3 ) in the above table), the merging processing is performed to update the target closed loop relationship graph, and step 205 is repeated until the target closed loop relationship graph no longer changes. Fifthly, for the element not included in the target closed loop relationship graph, the relationship between the above element and the elements in the existing target closed loop relationship graph is traversed. If there are more than or equal to two associated elements (for example, the relationship chain 0-1-2 in Figure 3 ) in the above table), the merging processing is performed to update the target closed loop relationship graph, and step 205 is repeated until the target closed loop relationship graph no longer changes.

[0058] Step 206, by traversing any two nodes in the target closed loop relationship graph, determining the shortest distance between the two nodes, and constructing an adjacency matrix based on the shortest distance.

[0059] Step 207, according to the adjacency matrix, calculating the shortest distance sum of each node in the target closed loop relationship graph.

[0060] Step 208, taking the node with the shortest distance sum satisfying the set condition as the key member in the target insurance fraud gang.

[0061] In the implementation, the shortest distance is the shortest path distance, the shortest path between two nodes is the shortest path that realizes the association between the two nodes, and the shortest path distance between two nodes is the length of the shortest path between the two nodes. For example, Figure 3As shown, taking the target closed-loop relationship graph h1 as an example, it contains a loop relationship point (0-5-4-6), and node (element) 1, node 2 and node 3 are connected to the nodes on the loop relationship. Starting from any vertex, the shortest distance from the vertex to all nodes (other nodes) in the target closed-loop relationship graph is calculated, the shortest distance sum is obtained, and at least one node with the shortest distance sum is taken as the key personnel of the gang. Specifically as follows:

[0062] Step 2081, assuming that the distance between two nodes is 1, then the distance between the two nodes is the shortest path distance, and using the breadth-first traversal method, it can be known that the path between node 0 and node 4 includes three: 0-6-4, the distance is 2; 0-5-3-4, the distance is 3; 0-5-4, the distance is 2, then the shortest path distance between node 0 and node 4 is 2.

[0063] Step 2082, taking different nodes in the target closed-loop relationship graph as vertices, the shortest distance to all nodes (other nodes) is calculated, as shown in Table 1, the shortest distance from any node to all nodes (other nodes) is obtained, as shown in Table 2, and the shortest distance sum of each node in the target closed-loop relationship graph is obtained by summing the adjacency matrix by column.

[0064] Table 1: Calculate the shortest distance sum

[0065] Point 0 Point 1 Point 2 Point 3 Point 4 Point 5 Point 6 Point 0 0 1 1 2 2 1 1 Point 1 1 0 1 3 3 2 2 Point 2 1 1 0 3 3 2 2 Point 3 2 3 3 0 1 1 2 Point 4 3 4 3 1 0 1 1 Point 5 1 2 2 1 1 0 2 Point 6 1 2 2 2 1 2 0

[0066] Table 2: Shortest distance sum of each node in the target closed-loop relationship graph

[0067] Point 0 Point 1 Point 2 Point 3 Point 4 Point 5 Point 6 Shortest distance and 9 13 12 12 11 9 10

[0068] It can be seen that based on the sorting of the shortest distance sum, the key degree of each node is sorted as node 0 (node 5) > node 6 > node 4 > node 2 (node 3) > node 1, based on the social scene, the personnel on the node with the smallest shortest distance sum can contact other personnel with less contact cost, and the personnel on the node with the smallest shortest distance sum is more important in the social network, therefore, node 0 and node 5 are considered as the key members of the target insurance fraud gang.

[0069] According to the requirements of the actual application scene, the breadth-first traversal method can use a auxiliary queue, any node is taken as a vertex and is enqueued, and is marked as visited, then it is looped to detect whether the queue is empty, if the queue is not empty, the first element in the queue is selected, all unvisited nodes associated with the element are enqueued, and these nodes are marked as visited; if the queue is empty, it means that all nodes have been traversed according to the breadth-first traversal.

[0070] By applying the technical scheme of the embodiment, based on the role information and the subject information in the newly added claim settlement event, a social network relationship graph is constructed using the insurance business relationship; according to the social network relationship graph, a traversal starting point for constructing a closed-loop relationship graph is obtained; using a first neighbor point acquisition method based on the claim settlement event relationship, a plurality of closed-loop relationship graphs for a plurality of traversal starting points are constructed; using a second neighbor point acquisition method based on the claim settlement event relationship, a target closed-loop relationship graph is obtained by merging the plurality of closed-loop relationship graphs, so as to obtain a technical scheme for finding a fraud gang, and further finding a key member by calculating the shortest path. Compared with the existing technical scheme based on the regular rule method, which can only identify fraud cases with key features and has low timeliness, and the machine learning method, which has high requirements for sample data and has low fraud case identification accuracy, it can be seen that by periodically traversing the closed-loop relationship of different insurance roles and insurance subjects in the shortest path of the insurance underwriting claim settlement event, the key members in the target insurance fraud gang can be effectively found.

[0071] Further, as Figure 1 a specific implementation of the method, the embodiment of the present application provides an insurance fraud gang identification system based on a social network, as Figure 4 shown, the system comprises a first construction module 41, a traversal module 42, a second construction module 43, and a merging module 44.

[0072] The first construction module 41 is configured to construct a social network relationship graph using the insurance business relationship based on the role information and the subject information in the newly added claim settlement event.

[0073] The traversal module 42 is configured to obtain a traversal starting point for constructing a closed-loop relationship graph according to the social network relationship graph.

[0074] The second construction module 43 is configured to construct a plurality of closed-loop relationship graphs for a plurality of traversal starting points using a first neighbor point acquisition method based on the claim settlement event relationship.

[0075] The merging module 44 is configured to obtain a target closed-loop relationship graph by merging the plurality of closed-loop relationship graphs using a second neighbor point acquisition method based on the claim settlement event relationship, and the node information in the target closed-loop relationship graph is used to represent a target insurance fraud gang.

[0076] In a specific application scenario, as Figure 5 shown, the first construction module 41 comprises a setting unit 411, a storage unit 412, and a connection unit 413.

[0077] The setting unit 411 is configured to take the role information in the newly added claim settlement event and the vehicle information in the subject information as nodes in the social network relationship graph.

[0078] The storage unit 412 is configured to obtain, for each node, a number of claim settlement events associated with the node, and store the number of claim settlement events on the node.

[0079] The connection unit 413 is configured to connect the nodes by using an insurance business relationship, and construct a social network relationship graph.

[0080] In a specific application scenario, the social network relationship graph includes: a first social network relationship graph constructed by traversing the new claim settlement event, the first social network relationship graph including the new claim settlement event; or / and, a second social network relationship graph constructed by traversing the historical claim settlement event and the new claim settlement event, the second social network relationship graph including the new claim settlement event and the historical claim settlement event.

[0081] In a specific application scenario, the target information includes a claim settlement amount corresponding to each claim settlement event, and the traversal module 42 is specifically configured to obtain a node corresponding to a maximum value of the number of claim settlement events in the social network relationship graph, and use the node as a traversal starting point for constructing a closed-loop relationship graph; or obtain a plurality of nodes corresponding to the maximum value of the number of claim settlement events in the social network relationship graph, and determine a traversal starting point for constructing a closed-loop relationship graph according to claim settlement amount information in the plurality of nodes.

[0082] In a specific application scenario, the second construction module 43 includes a first obtaining unit 431 and a second obtaining unit 432.

[0083] The first obtaining unit 431 is configured to construct a relationship chain based on a current traversal starting point by using a first neighbor point obtaining manner based on a claim settlement event relationship; wherein the first neighbor point obtaining manner is to obtain a plurality of nodes having the same insurance business relationship with the traversal starting point, and determine a direct neighbor point of the traversal starting point according to a number of claim settlement events corresponding to the same insurance business relationship; the direct neighbor point is a neighbor point between the traversal starting point and the neighbor point without other nodes, and the indirect neighbor point is a neighbor point between the traversal starting point and the neighbor point with other nodes.

[0084] The second obtaining unit 432 is configured to obtain an indirect neighbor point having the same insurance business relationship with the traversal starting point in the relationship chain, and construct a plurality of closed-loop relationship graphs for the current traversal starting point based on the indirect neighbor point.

[0085] In a specific application scenario, the merging module 44 is specifically configured to, if there is an associated node having the same claim settlement event relationship in at least two closed-loop relationship graphs in the plurality of closed-loop relationship graphs, perform merging processing on the at least two closed-loop relationship graphs based on the associated node, and obtain a target closed-loop relationship graph.

[0086] In a specific application scenario, the system further comprises a determination module 45.

[0087] The determination module 45 is configured to determine the shortest distance between any two nodes in the target closed-loop relationship graph by traversing the two nodes, construct an adjacency matrix based on the shortest distance, calculate the shortest distance sum of each node in the target closed-loop relationship graph according to the adjacency matrix, and take a node with a shortest distance sum satisfying a set condition as a key member in the target insurance fraud gang.

[0088] It should be noted that other corresponding descriptions of the various functional units involved in the insurance fraud gang identification system based on a social network provided by the embodiments of the present application can be referred to the corresponding descriptions in Figure 1 and Figure 2 , which will not be described here in detail.

[0089] Based on the above method as shown in Figure 1 and Figure 2 , accordingly, the embodiments of the present application also provide a computer storage medium having a computer program stored thereon, which is executed by a processor to implement the above-mentioned insurance fraud gang identification method based on a social network as shown in Figure 1 and Figure 2 .

[0090] Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which can be stored in a storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.) and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in various implementation scenarios of the present application.

[0091] Based on the above method as shown in Figure 1 , Figure 2 , and the virtual system embodiment as shown in Figure 4 , Figure 5 , in order to achieve the above-mentioned purpose, the embodiments of the present application further provide a computer device, which can be a personal computer, a server, a network device, etc. The physical device includes a storage medium and a processor; the storage medium is configured to store a computer program; and the processor is configured to execute the computer program to implement the above-mentioned insurance fraud gang identification method based on a social network as shown in Figure 1 and Figure 2 .

[0092] Optionally, the computer device can further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, and the like. The user interface can include a display screen, an input unit such as a keyboard, and the like. Optionally, the user interface can further include a USB interface, a card reader interface, and the like. The network interface can optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface), and the like.

[0093] Those skilled in the art can understand that the computer device structure provided by the embodiment does not constitute a limitation on the entity device, and can include more or fewer components, or combine certain components, or different component arrangements.

[0094] The storage medium can further include an operating system, a network communication module. The operating system is a program for managing hardware and software resources of the computer device, supporting the running of information processing programs and other software and / or programs. The network communication module is used to realize communication between components in the storage medium, and communication with other hardware and software in the entity device.

[0095] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary general hardware platforms, or by hardware. By applying the technical solutions of the present application, compared with the existing regular rule method which can only identify fraud cases with key features, the solution has lower timeliness, and the machine learning method has higher requirements for sample data, so there is a lower accuracy of fraud case identification. Compared with the technical solutions, based on the role information and the subject information in the newly added claim settlement event, a social network relationship graph is constructed by using the insurance business relationship; according to the social network relationship graph, a traversal starting point for constructing a closed-loop relationship graph is obtained; a plurality of closed-loop relationship graphs for a plurality of traversal starting points are constructed by using a first neighbor point acquisition method based on the claim settlement event relationship; a target closed-loop relationship graph is obtained by merging processing the plurality of closed-loop relationship graphs by using a second neighbor point acquisition method based on the claim settlement event relationship, so as to obtain a technical solution for finding a fraud gang, and further finding a key member by calculating a shortest path. It can be seen that by periodically traversing the closed-loop relationship of the correlation relationship between different insurance roles and insurance subjects in the shortest path in the insurance claim settlement event, the key members in the target insurance fraud gang can be effectively found.

[0096] Those skilled in the art can understand that the modules or flows in the drawings are not necessarily required for implementing the present application. Those skilled in the art can understand that the modules in the system in the implementation scenario can be distributed in the system in the implementation scenario according to the description of the implementation scenario, or can be changed to be located in one or more systems different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple units.

[0097] The above application number is only for description, and does not represent the advantages and disadvantages of the implementation scenario. The above disclosure is only some specific implementation scenarios of the present application, but the present application is not limited thereto, and any variations that can be thought of by those skilled in the art should fall within the protection scope of the present application.

Claims

1. A social network-based insurance fraud ring identification method, characterized in that, include: Based on the role and object information in newly added claims events, a social network relationship graph is constructed using insurance business relationships; Based on the social network relationship graph, obtain the traversal starting point for constructing the closed-loop relationship graph; Using a first neighbor acquisition method based on claims event relationships, multiple closed-loop relationship graphs are constructed for multiple traversal starting points. The first neighbor acquisition method involves acquiring multiple nodes with the same insurance business relationship as the traversal starting point, and determining the direct neighbors of the traversal starting point based on the number of claims events corresponding to the same insurance business relationship. Using the first neighbor acquisition method based on claims event relationships, a relationship chain is constructed based on the current traversal starting point. Indirect neighbors in the relationship chain with the same insurance business relationship as the traversal starting point are acquired, and multiple closed-loop relationship graphs are constructed based on these indirect neighbors. A direct neighbor is one where there are no other nodes between the traversal starting point and its neighbor, while an indirect neighbor is one where there are other nodes between the traversal starting point and its neighbor. By using a second neighbor node acquisition method based on claims event relationships, a target closed-loop relationship graph is obtained by merging the multiple closed-loop relationship graphs. The node information in the target closed-loop relationship graph is used to characterize the target insurance fraud gang. If at least two of the multiple closed-loop relationship graphs have associated nodes with the same claims event relationship, the at least two closed-loop relationship graphs are merged based on the associated nodes to obtain the target closed-loop relationship graph. The method further includes: By traversing any two nodes in the target closed-loop relationship graph, the shortest distance between any two nodes is determined, and an adjacency matrix is ​​constructed based on the shortest distance. Based on the adjacency matrix, the sum of the shortest distances for each node in the target closed-loop graph is calculated; The nodes with the shortest distance and those that meet the set conditions are considered key members of the target insurance fraud gang.

2. The method of claim 1, wherein, The social network diagram includes: A first social network graph is constructed by traversing the newly added claim events. This first social network graph includes the newly added claim events; or / and... A second social network graph is constructed by traversing historical claim events and newly added claim events. The second social network graph includes newly added claim events and historical claim events.

3. The method according to claim 1, characterized in that, The steps for constructing a social network graph based on role and insured information in newly added claims events, utilizing insurance business relationships, include: The role information in the newly added claims event and the vehicle information in the target information are used as nodes in the social network relationship graph; For each node, obtain the number of claim events associated with the node and store the number of claim events on the node; By connecting the nodes using insurance business relationships, a social network relationship graph is constructed.

4. The method according to claim 3, characterized in that, The target information includes the claim amount corresponding to each claim event. The step of obtaining the traversal starting point for constructing the closed-loop relationship graph based on the social network relationship graph includes: Obtain the node corresponding to the maximum number of claim events in the social network graph, and use it as the starting point for traversing the closed-loop graph; or, Obtain multiple nodes corresponding to the maximum number of claim events in the social network relationship graph, and determine the traversal starting point for constructing the closed-loop relationship graph based on the claim amount information in the multiple nodes.

5. A social network-based insurance fraud gang identification system, characterized in that, include: The first construction module is used to build a social network relationship graph based on the role information and object information in new claims events, utilizing insurance business relationships. The traversal module is used to obtain the starting point for traversing the social network relationship graph to construct the closed-loop relationship graph. The second construction module is used to construct multiple closed-loop relationship graphs for multiple traversal starting points by utilizing the first neighbor point acquisition method based on the relationship between claims events. The second construction module includes a first acquisition unit and a second acquisition unit. The first acquisition unit is used to construct a relationship chain based on the current traversal starting point by using the first neighbor acquisition method based on the relationship of the claim event. The second acquisition unit is used to acquire indirect neighboring nodes in the relationship chain that have the same insurance business relationship as the traversal starting point, and to construct multiple closed-loop relationship graphs for the current traversal starting point based on the indirect neighboring nodes; wherein, the first neighboring node acquisition method is to acquire multiple nodes that have the same insurance business relationship as the traversal starting point, and to determine the direct neighboring nodes of the traversal starting point according to the number of claims events corresponding to the same insurance business relationship; the direct neighboring node is one where there are no other nodes between the traversal starting point and the neighboring node, and the indirect neighboring node is one where there are other nodes between the traversal starting point and the neighboring node; The merging module is used to obtain a target closed-loop relationship graph by merging the multiple closed-loop relationship graphs using the second neighbor point acquisition method based on the relationship of claims events. The node information in the target closed-loop relationship graph is used to characterize the target insurance fraud gang. The merging module is specifically used to merge the at least two closed-loop relationship graphs based on the associated nodes if there are related nodes with the same claim event relationship in at least two of the multiple closed-loop relationship graphs, to obtain the target closed-loop relationship graph. The determination module is used to determine the shortest distance between any two nodes in the target closed-loop relationship graph by traversing the graph, and to construct an adjacency matrix based on the shortest distance; to calculate the sum of the shortest distances of each node in the target closed-loop relationship graph according to the adjacency matrix; and to identify the nodes whose sum of the shortest distances meets the set conditions as key members of the target insurance fraud gang.

6. A computer storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the social network-based insurance fraud gang identification method according to any one of claims 1 to 4.

7. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the social network-based insurance fraud gang identification method according to any one of claims 1 to 4.

Images