Method and system for identifying risk business behavior of life insurance company

Abstract

The invention provides a life insurance company risk business behavior identification method and system, and the method comprises the steps: taking an insurance policy, a customer and a sales agent as nodes, taking an insurance relationship, a sales relationship and a loan relationship among the insurance policy, the customer and the sales agent as edges, and building a risk business behavior identification business map; identifying an insurance policy node, a sales agent node, a client node and edges among the insurance policy node, the sales agent node and the client node included in a sub-graph in the business graph according to the risk business behavior, and identifying a first risk business behavior; and identifying the second risk business behavior according to the insurance policy node, the client node and the edge between the insurance policy node and the client node included in the sub-graph in the business graph according to the risk business behavior, thereby effectively improving the risk business behavior identification efficiency and reliability.

Description

Technical Field

[0001] The present invention relates to the field of abnormal insurance behavior identification, and in particular to a method and system for identifying risky business behaviors of life insurance companies. Background Art

[0002] In the insurance business of life insurance companies, "splitting orders" refers to the fact that the policies sold by sales agents meet the "three sames but different" conditions (the same customer, in the same month, insured policies of the same type under different sales agents), and "circular insurance" refers to the fact that after a customer takes out insurance, he or she takes out a loan for the policy in a short period of time (the customer can apply for a loan based on the cash value of the effective policy), and then takes out a policy of the same type again after the loan, and the loan has not been repaid at the time of insurance. This can be simply understood as "taking money from Peter to pay Paul." The above two behaviors are abnormal sales behaviors caused by assessment pressure and arbitrage by the sales team, which are likely to generate false premiums, increase the company's sales costs, and cause losses and operating risks to the company.

[0003] Therefore, life insurance companies need to promptly detect "splitting orders" or "circular insurance" behaviors. Traditional identification methods are generally based on relational databases, and identify splitting orders or circular insurance behaviors through association analysis of multiple business database tables. Identifying "splitting orders" and "circular insurance" behaviors through a large number of table associations has high query costs, and it is difficult to ensure identification efficiency. In addition, the amount of coding required to implement this function is high, and it is difficult to quickly visualize the identification results, and it is impossible to intuitively understand the "risk" behavior, which makes the identification efficiency of splitting orders and / or circular insurance behaviors low.

[0004] In order to solve the problem of low efficiency in identifying order splitting behavior and / or circular insurance behavior in the prior art, this solution is proposed. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention innovatively proposes a method and system for identifying the risk business behavior of a life insurance company, which effectively solves the problem of low efficiency in identifying the bill splitting behavior and / or circular insurance behavior caused by the prior art, and effectively improves the efficiency in identifying the bill splitting behavior and / or circular insurance behavior.

[0006] A first aspect of the present invention provides a method for identifying risky business behaviors of a life insurance company, comprising:

[0007] A business graph for identifying risky business behaviors is established with insurance policies, customers, and sales agents as nodes, and with the insurance relationships, sales relationships, and loan relationships among insurance policies, customers, and sales agents as edges; wherein risky business behaviors include first risky business behaviors and / or second risky business behaviors, wherein the first risky business behaviors are splitting behaviors, and the second risky business behaviors are cyclic insurance behaviors;

[0008] Identify the first risky business behavior by identifying the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in the subgraph of the risky business behavior identification business graph; and / or,

[0009] According to the risk business behavior, the policy nodes and customer nodes included in the subgraph in the business graph and the edges between the policy nodes and the customer nodes are identified to identify the second risk business behavior.

[0010] Optionally, each customer node in the risk business behavior identification business graph represents a customer; each insurance policy node represents an insurance policy; each sales agent node represents a sales agent;

[0011] There is a first edge and a second edge between the customer node and the policy node, and there is a third edge between the policy node and the sales agent node; the first edge is from the customer node to the policy node, representing an insurance relationship; the second edge is from the customer node to the policy node, representing a policy loan relationship; the third edge is from the sales agent node to the policy node, representing a policy sales relationship; the policy node includes policy type attribute information.

[0012] Further, the first side is marked as an insurance type, the second side is marked as a loan type, and the third side is marked as a sales type;

[0013] The first edge includes first attribute information and second attribute information, the first attribute information is the insurance time, and the second attribute information is the insurance effective time;

[0014] The second side includes fourth attribute information and fifth attribute information, the fourth attribute information is the policy loan time, and the fifth attribute information is the repayment time of the policy loan.

[0015] Further, according to the risk business behavior identification business graph, the policy node, the sales agent node, the customer node, and the edges between the policy node, the sales agent node, and the customer node are included in the subgraph, and the first risk business behavior is specifically identified as follows:

[0016] If the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in a subgraph in the risk business behavior identification business graph meet the preset first risk identification rules, then the insurance behavior involved in the subgraph constitutes a first risk business behavior.

[0017] Furthermore, the first risk identification rule includes:

[0018] The same customer node is connected to at least two policy nodes; and,

[0019] The first type of policy node is connected to at least two different sales agent nodes, wherein the first type of policy node is at least two policy nodes connected to the same customer node; and,

[0020] The first attribute information or the second attribute information of the first edge between the same customer node and the first type of policy node is in the same year and month; and,

[0021] The policy type attribute information of the first type of policy nodes are all the same.

[0022] Optionally, identifying the second risky business behavior specifically includes: identifying the policy node, the customer node, and the edge between the policy node and the customer node included in the subgraph in the risky business behavior identification business graph;

[0023] If the policy nodes, customer nodes, and edges between policy nodes and customer nodes in the risk business behavior identification business graph meet the preset second risk identification rules, then the insurance behavior involved in the subgraph constitutes a second risk business behavior.

[0024] Furthermore, the second risk identification rule includes:

[0025] The same customer node corresponds to at least two policy nodes, and there is a second edge between at least one of the first-type policy nodes and the customer node, wherein the first-type policy nodes are at least two policy nodes corresponding to and connected to the same customer node; and,

[0026] The first attribute information of the first edge of the first category is less than the fourth attribute information of the second edge of the first category, wherein the first edge of the first category is the first edge between the same customer node and the policy node having a second edge with the customer node; the second edge of the first category is the second edge between the same customer node and the policy node having a second edge with the customer node; and,

[0027] The fourth attribute information of the first-category second edge is less than the first attribute information of the second-category first edge, wherein the second-category first edge is the first edge between the same customer node and the policy node between which no second edge exists; and,

[0028] The fifth attribute information of the second edge of the first category is greater than the first attribute information of the first edge of the second category; and,

[0029] The policy type attribute information of the first type of policy nodes are all the same.

[0030] Optionally, the more sales agents included in each subgraph involved in the first risk business behavior, or the more policy nodes corresponding to each sales agent in each subgraph involved in the first risk business, the higher the risk level of the first risk business behavior corresponding to the subgraph;

[0031] The more policy nodes corresponding to the same customer node included in each subgraph involved in the second risk business behavior, or the more policy nodes that have a second edge with the same customer node included in each subgraph involved in the second risk business, the higher the risk level of the second risk business behavior.

[0032] Optionally, it also includes:

[0033] Marking the customer nodes, policy nodes, sales agent nodes included in each subgraph involved in the first risk business behavior and the edges between the customer nodes, policy nodes, and sales agent nodes involved in the first risk business; and / or,

[0034] The customer nodes and policy nodes included in each subgraph involved in the second risk business behavior and the edges between the customer nodes and policy nodes involved in the second risk business are marked.

[0035] A second aspect of the present invention provides a system for identifying risky business behaviors of a life insurance company, comprising:

[0036] A module is established, with insurance policies, customers, and sales agents as nodes, and the insurance relationship, sales relationship, and loan relationship between insurance policies, customers, and sales agents as edges, to establish a business graph for identifying risky business behaviors; wherein risky businesses include first risky businesses and second risky businesses, wherein the first risky business is the splitting behavior, and the second risky business is the circular insurance behavior;

[0037] A first identification module identifies the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in the subgraph of the business graph according to the risk business behavior, and identifies the first risk business; and / or

[0038] The second identification module identifies the policy nodes, customer nodes, and edges between the policy nodes and customer nodes included in the subgraph in the business graph based on the risk business behavior, and identifies the second risk business.

[0039] The technical solution adopted by the present invention includes the following technical effects:

[0040] 1. In the technical solution of the present invention, a business graph for identifying risky business behaviors is established with insurance policies, customers, and sales agents as nodes, and with insurance relationships, sales relationships, and loan relationships among insurance policies, customers, and sales agents as edges; a first risky business behavior is identified based on the insurance policy nodes, sales agent nodes, and customer nodes included in the subgraph of the risky business behavior identification business graph, and the edges between the insurance policy nodes, sales agent nodes, and customer nodes; a second risky business behavior is identified based on the insurance policy nodes, customer nodes, and the edges between the insurance policy nodes and customer nodes included in the subgraph of the risky business behavior identification business graph, thereby effectively solving the problem of low efficiency in identifying order splitting behaviors and / or circular insurance behaviors caused by the prior art, and effectively improving the efficiency in identifying order splitting behaviors and / or circular insurance behaviors.

[0041] 2. The technical solution of the present invention abandons the traditional data modeling method based on relational databases, and constructs a business graph dedicated to the identification of "splitting orders" and "cyclic insurance" behaviors in the graph database based on the "graph model". Through clever data modeling and combining the advantages of graph databases in graph data analysis and mining, the "sub-graph identification" method is used to efficiently identify the behaviors of splitting orders and cyclic insurance, which reduces the difficulty of identification, improves the identification efficiency, and improves the accuracy and comprehensiveness of the identification results.

[0042] 3. In the technical solution of the present invention, the customer nodes, policy nodes, sales agent nodes included in each subgraph involved in the first risk business and the edges between the customer nodes, policy nodes and sales agent nodes involved in the first risk business are marked; the customer nodes, policy nodes included in each subgraph involved in the second risk business and the edges between the customer nodes and policy nodes involved in the second risk business are marked, and the identification return result is still graph data (subgraph), which is easy to visualize and can fully deconstruct the various parties involved in the "order splitting" and circular insurance behavior.

[0043] 4. In the technical solution of the present invention, the more sales agents included in each subgraph involved in the first risk business behavior, or the more policy nodes corresponding to each sales agent in each subgraph involved in the first risk business, the higher the risk level of the first risk business behavior corresponding to the subgraph; the more policy nodes corresponding to the same customer node included in each subgraph involved in the second risk business behavior, or the more policy nodes with a second edge with the same customer node included in each subgraph involved in the second risk business, the higher the risk level of the second risk business behavior. It can quickly identify subgraphs with higher risks, reduce the company's losses and risks of fraud, and improve business quality. It can help life insurance companies efficiently identify "splitting orders" and "circular insurance" behaviors, fully "understand" the various representative behavior patterns involved in the "splitting orders" of sales agents and the "circular insurance behavior" of customers, and pay special attention to sales agents and customers who frequently engage in the above-mentioned risky business behaviors, provide risk reminders and warnings, guide business behaviors to be more standardized, and help the company develop high-quality development.

[0044] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 A schematic diagram of a process of a method in Embodiment 1 of the present invention;

[0047] Figure 2 This is a schematic diagram illustrating the "splitting orders" and "circular insurance" behaviors in the method of Example 1 of the solution of the present invention;

[0048] Figure 3 This is an example schematic diagram of a business graph for identifying risky business behaviors in the method of Embodiment 1 of the solution of the present invention;

[0049] Figure 4 This is an example schematic diagram of sub-graph identification in a risk business behavior identification business graph in the method of Embodiment 1 of the solution of the present invention;

[0050] Figure 5 This is an example schematic diagram of order splitting behavior mode 1 in the method of embodiment 1 of the solution of the present invention;

[0051] Figure 6 This is an example schematic diagram of order splitting behavior mode 2 in the method of embodiment 1 of the solution of the present invention;

[0052] Figure 7 This is an example schematic diagram of mode 3 of the order splitting behavior (first risk business behavior) in the method of embodiment 1 of the solution of the present invention;

[0053] Figure 8 This is an example schematic diagram of order splitting behavior mode 4 in the method of embodiment 1 of the solution of the present invention;

[0054] Fig. 9 This is an example schematic diagram of the cyclic insurance behavior (second risk business behavior) mode 1 in the method of embodiment 1 of the solution of the present invention;

[0055] Fig.10 This is an example schematic diagram of the cyclic insurance behavior mode 2 in the method of embodiment 1 of the present invention;

[0056] Fig.11 Another schematic diagram of the process of the method of Embodiment 1 in the scheme of the present invention;

[0057] Fig.12 It is a schematic diagram of the structure of the system of Example 2 in the solution of the present invention. DETAILED DESCRIPTION

[0058] In order to clearly illustrate the technical features of the present solution, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings. The disclosure below provides many different embodiments or examples for realizing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and settings of specific examples are described below. In addition, the present invention may repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. It should be noted that the components illustrated in the accompanying drawings are not necessarily drawn to scale. The present invention omits the description of known components and processing techniques and processes to avoid unnecessary limitations on the present invention.

[0059] Embodiment 1

[0060] like Figure 1 As shown, the present invention provides a method for identifying risky business behaviors of a life insurance company, comprising:

[0061] S1, with the insurance policy, customer, and sales agent as nodes, and the insurance relationship, sales relationship, and loan relationship between the insurance policy, customer, and sales agent as edges, a business graph for identifying risky business behaviors is established; wherein the risky business behaviors include the first risky business behaviors and / or the second risky business behaviors, the first risky business behaviors are the splitting behaviors, and the second risky business behaviors are the circular insurance behaviors;

[0062] S2, identifying the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in the subgraph of the business graph according to the risk business behavior, and identifying the first risk business behavior; and / or,

[0063] S3, identifying the policy nodes, customer nodes, and edges between the policy nodes and customer nodes included in the subgraph in the business graph according to the risky business behavior, and identifying the second risky business behavior.

[0064] Among them, in step S1, if Figure 2-Figure 3 As shown in Figure 2, both the “splitting order” behavior recognition and the “circular insurance” behavior recognition require analyzing the business relationship between the insurance policy, the sales agent, and the customer. Figure 2 As shown in Figure 2, this project builds a dedicated graph model based on the principle of maximizing the ability to identify "order splitting" and "circular insurance" behaviors, as shown in Figure 2. Figure 3 As shown, the graph contains all the element information required for identifying the two risk behaviors.

[0065] In the risk business behavior identification business graph, each customer node represents a customer; each policy node represents an insurance policy; and each sales agent node represents a sales agent.

[0066] There is a first edge and a second edge between the customer node and the policy node, and there is a third edge between the policy node and the sales agent node; the first edge points from the customer node to the policy node, representing an insurance relationship; the second edge points from the customer node to the policy node, representing a policy loan relationship; the third edge points from the sales agent node to the policy node, representing a policy sales relationship; the policy node includes policy type attribute information.

[0067] The first side is labeled as insurance type, the second side is labeled as loan type, and the third side is labeled as sales type;

[0068] The first side includes first attribute information, second attribute information, and third attribute information, the first attribute information is the insurance time, the second attribute information is the insurance effective time (or effective time), and the third attribute information is the insurance effective premium (or effective premium);

[0069] The second side includes fourth attribute information, fifth attribute information, and sixth attribute information, wherein the fourth attribute information is the policy loan time (or loan time), the fifth attribute information is the policy loan repayment time (or repayment time), and the sixth attribute information is the policy loan amount (or loan amount);

[0070] The third edge includes seventh attribute information, eighth attribute information, and ninth attribute information. The seventh attribute information is the insurance time, the eighth attribute information is the insurance effective time (or effective time), and the ninth attribute information is the insurance effective premium (or effective premium), which can facilitate management and maintenance personnel to timely understand the attribute information of the nodes and edges involved in the first risk business behavior or the second risk business behavior.

[0071] That is, there are three types of points in the risk business behavior identification business graph: policy, customer, and sales agent, which represent an insurance policy, a customer, and a sales agent respectively.

[0072] The first edge of the "insurance" type is used to connect the node of the "customer" type and the node of the "policy" type. The first edge points from the point of the "customer" type to the point of the "policy" type, representing the customer's insurance behavior for the policy; the third edge of the "sales" type is used to connect the node of the "sales agent" type and the node of the "policy" type. The third edge points from the point of the "sales agent" type to the node of the "policy" type, representing the sales agent's sales behavior for the policy; the second edge of the "borrowing" type is used to connect the node of the "customer" type and the node of the "policy" type. The second edge points from the node of the "customer" type to the node of the "policy" type, representing the customer's borrowing behavior for the policy;

[0073] The first side of the "insurance" type includes three attributes of the policy: "insurance time", "insurance effective time", and "insurance effective premium", namely, first attribute information, second attribute information, and third attribute information.

[0074] The second side of the "loan" type contains three attributes: the customer's "policy loan time", the customer's "policy loan repayment time", and the customer's "policy loan amount", namely the fourth attribute information, the fifth attribute information, and the sixth attribute information.

[0075] The third side of the "sales" type contains three attributes of the policy: "insurance time", "effective time" and "insurance effective premium" of the policy, namely the seventh attribute information, the eighth attribute information and the ninth attribute information.

[0076] The node of type "insurance policy" contains the insurance type attributes of the insurance policy.

[0077] Among them, in step S2, according to the risk business behavior, the policy node, the sales agent node, the customer node, and the edges between the policy node, the sales agent node, and the customer node included in the subgraph in the business graph are identified, and the identification of the first risk business behavior specifically includes:

[0078] If the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in a subgraph in the risk business behavior identification business graph meet the preset first risk identification rules, then the insurance behavior involved in the subgraph constitutes a first risk business behavior.

[0079] Specifically, the first risk identification rule includes:

[0080] The same customer node is connected to at least two policy nodes; and,

[0081] The first type of policy node is correspondingly connected to at least two different sales agent nodes, wherein the first type of policy node is at least two policy nodes correspondingly connected to the same customer node (i.e., different policy nodes connected to the same customer node); and,

[0082] The first attribute information or the second attribute information of the first edge between the same customer node and the first type of policy node is in the same year and month; and,

[0083] The policy type attribute information of the first type of policy nodes are all the same.

[0084] Specifically, compared with the relational database modeling method, the graph model has strong business rule expression capabilities and is suitable for analyzing complex relationships between behavioral entities. Reflected in the graph, the "order splitting" behavior is a specific subgraph pattern (graph pattern: the basic unit that describes the graph structure, including the definition of points and edges and the relationship type, edge type, edge direction, and point and edge attributes between them). The relationship between "sales agent" and "insurance policy" that "conforms" to the order splitting behavior rules can be discovered through "subgraph identification", such as Figure 4 As shown, the nodes and edges with bold lines constitute a subgraph identified as the first business risk behavior or the second business risk behavior.

[0085] This method does not depend on the specific graph database type. Taking Nebula graph database as an example, based on the business graph constructed above, the following nGQL (NebulaGraph graph database declarative query language, compatible with most opencypher syntax) can be used to quickly identify subgraphs that meet the "order splitting" behavior pattern. The expression is natural and the implementation method is as follows:

[0086] MATCH p=(a:agent)-[:sales]->(b:policy)<-[f:insurance]-(e:customer)-[g:insurance]->(c:policy)<-[:sales]-(d:agent)

[0087] WHERE f.Insurance time.month == g.Insurance time.month

[0088] AND f.Insurance time.year == g.Insurance time.year

[0089] AND b!=c

[0090] AND b.Policy.Insurance Type == c.Policy.Insurance Type

[0091] AND a!=d

[0092] RETURN p limit 100 (for smooth and fast execution);

[0093] Take the "splitting order" behavior identification as an example:

[0094] like Figure 5 As shown, the splitting behavior in splitting behavior mode 1 involves two agents, and each agent has only one policy involved in splitting; the customer node (Wang xx) is connected to two policy nodes (policy 1 and policy 2); and,

[0095] The first type of policy nodes (Policy 1 and Policy 2) are connected to two different sales agent nodes (Ai xx, Li xx); and,

[0096] The first attribute information (insurance time) or the second attribute information (insurance effective time) of the first edge between the same customer node and the first type of insurance policy node are the same in year and month; and the insurance policy type attribute information of the first type of insurance policy nodes are the same.

[0097] like Figure 6 As shown, the splitting behavior in splitting behavior mode 2 involves two agents, and there are two or more policies involving splitting by the agent; the customer node (Guo xx) is connected to 4 policy nodes (policy 1, policy 2, policy 3, policy 4); and,

[0098] The first type of policy nodes (Policy 1, Policy 2, Policy 3, Policy 4) are connected to two different sales agent nodes (Teng xx, Cao xx); and,

[0099] The first attribute information (insurance time) or the second attribute information (insurance effective time) of the first edge between the same customer node and the first type of insurance policy node are the same in year and month; and the insurance policy type attribute information of the first type of insurance policy nodes are the same.

[0100] like Figure 7 As shown, the splitting behavior in splitting behavior mode 3 involves three or more agents, and there are two or more policies involving splitting by the agent; the customer node (Wang xx) is connected to 4 policy nodes (policy 1, policy 2, policy 3, policy 4); and,

[0101] The first type of policy nodes (policy 1, policy 2, policy 3, policy 4) are connected to three different sales agent nodes (Li xx, Huang xx, Li x); and,

[0102] The first attribute information (insurance time) or the second attribute information (insurance effective time) of the first edge between the same customer node and the first type of insurance policy node are the same in year and month; and the insurance policy type attribute information of the first type of insurance policy nodes are the same.

[0103] like Figure 8 As shown, the splitting behavior in splitting behavior mode 4 involves three or more agents, and there are two or more policies involving splitting by the agent; the customer node (Wang xx) is connected to 6 policy nodes (policy 1, policy 2, policy 3, policy 4, policy 5, policy 6); and,

[0104] The first type of policy nodes (policy 1, policy 2, policy 3, policy 4, policy 5, policy 6) are connected to three different sales agent nodes (Li xx, Jiang xx, Han xx); and,

[0105] The first attribute information (insurance time) or the second attribute information (insurance effective time) of the first edge between the same customer node and the first type of insurance policy node are the same in year and month; and the insurance policy type attribute information of the first type of insurance policy nodes are the same.

[0106] As the complexity of order splitting behavior patterns continues to increase, business risks are also getting higher and higher (the risk levels of order splitting behavior pattern 1, order splitting behavior pattern 2, order splitting behavior pattern 3, and order splitting behavior pattern 4 increase in sequence), which fully proves the practicality and reliability of this identification solution in the field of "order splitting" behavior identification.

[0107] That is, the more sales agents included in each subgraph involved in the first risk business behavior, or the more policy nodes corresponding to each sales agent in each subgraph involved in the first risk business, the higher the risk level of the first risk business behavior corresponding to the subgraph.

[0108] Among them, in step S3, according to the risk business behavior, the policy node, the customer node, and the edge between the policy node and the customer node included in the subgraph in the business graph are identified, and the second risk business behavior is specifically identified as follows:

[0109] If the policy nodes, customer nodes, and edges between policy nodes and customer nodes in the risk business behavior identification business graph meet the preset second risk identification rules, then the insurance behavior involved in the subgraph constitutes a second risk business behavior.

[0110] Specifically, the second risk identification rule includes:

[0111] At least two policy nodes correspond to the same customer node, and there is a second edge between at least one policy node in the first type of policy nodes and the customer node, where the first type of policy nodes are at least two policy nodes corresponding to the same customer node; and,

[0112] The first attribute information of the first type of first edges is all less than the fourth attribute information of the first type of second edges, where the first type of first edges are the first edges between the same customer node and the policy nodes with a second edge between the customer node and the policy nodes (the first edges between the customer node with a second edge and the policy nodes); the first type of second edges are the second edges between the same customer node and the policy nodes with a second edge between the customer node and the policy nodes (the second edges between the customer node with a second edge and the policy nodes); and,

[0113] The fourth attribute information of the first type of second edges is all less than the first attribute information of the second type of first edges, where the second type of first edges are the first edges between the same customer node and the policy nodes without a second edge between the customer node and the policy nodes (the first edges between the customer node without a second edge and the policy nodes); and,

[0114] The fifth attribute information of the first type of second edges is all greater than the first attribute information of the second type of first edges; and,

[0115] The policy type attribute information of the first type of policy nodes is all the same.

[0116] Specifically, compared with the relational database modeling method, the graph model has strong business rule expression ability and is suitable for analyzing complex associations between actors. Reflected in the graph, the "circular insurance application" behavior is a specific subgraph pattern (graph pattern: the basic unit describing the graph structure, including the definitions of nodes and edges and the relationship types, edge types, edge directions, and attributes of nodes and edges) between "customers" and "policies" that conform to the "circular insurance application" behavior rule can be discovered through "subgraph recognition", such as Figure 4 shown.

[0117] This method does not depend on the specific graph database type. Taking the Nebula graph database as an example, based on the above constructed business graph, the following nGQL (NebulaGraph graph database declarative query language, compatible with most opencypher syntax) can be used to quickly identify the subgraph that conforms to the "circular insurance application" behavior pattern. The expression is natural, and the specific implementation method is as follows:

[0118] MATCH p=(c:policy)<-[g:insured]-(b:customer)-[f:borrow]->(a:policy)<-[e:insured]-(b:customer)

[0119] WHERE e.insured_time < f.borrow_time

[0120] AND f. Borrowing time < g. Insurance application time

[0121] AND f. Repayment time > g. Insurance application time

[0122] AND a!= c

[0123] AND a. Policy. Insurance type == c. Policy. Insurance type

[0124] RETURN p LIMIT 100;

[0125] As Fig. 9 shown, the same customer node (Xu xx) corresponds to two policy nodes (Policy 1 and Policy 2), and there is a second edge (loan) between the policy node (Policy 1) in the first type of policy nodes (Policy 1 and Policy 2) and this customer node; and,

[0126] The first attribute information (insurance application time) of the first edge in the first type is less than the fourth attribute information (borrowing time) of the second edge in the first type. Among them, the first edge in the first type is the first edge between the same customer node (Xu xx) and the policy node (Policy 1) that has a second edge with this customer node; the second edge in the first type is the second edge between the same customer node (Xu xx) and the policy node (Policy 1) that has a second edge with this customer node; and,

[0127] The fourth attribute information (policy loan time) of the second edge in the first type is less than the first attribute information (insurance application time) of the first edge in the second type. Among them, the first edge in the second type is the first edge between the same customer node (Xu xx) and the policy node (Policy 2) that has no second edge with this customer node; and,

[0128] The fifth attribute information (repayment time of policy loan) of the second edge in the first type is greater than the first attribute information (insurance application time) of the first edge in the second type; and, the policy insurance type attribute information of the first type of policy nodes is the same.

[0129] As Fig.10 shown, the same customer node (Xu xx) corresponds to four policy nodes (Policy 1, Policy 2, Policy 3, Policy 4), and there is a second edge (loan) between the policy nodes (Policy 1, Policy 3, Policy 4) in the first type of policy nodes (Policy 1, Policy 2, Policy 3, Policy 4) and this customer node; and,

[0130] The first attribute information (insurance time) of the first edge of the first category is less than the fourth attribute information (loan time) of the second edge of the first category, wherein the first edge of the first category is the first edge between the same customer node (Xu xx) and the policy nodes (Policy 1, Policy 3, Policy 4) with the second edge between the customer node; the second edge of the first category is the second edge between the same customer node (Xu xx) and the policy nodes (Policy 1, Policy 3, Policy 4) with the second edge between the customer node; and,

[0131] The fourth attribute information (policy loan time) of the second edge of the first category is less than the first attribute information (insurance time) of the first edge of the second category, where the first edge of the second category is the first edge between the same customer node (Xu xx) and the policy nodes (policy 2, policy 5) that do not have a second edge between the customer node and the policy node; and,

[0132] The fifth attribute information (repayment time of the policy loan) of the second edge of the first category is greater than the first attribute information (insurance time) of the first edge of the second category; and the policy type attribute information of the first category policy nodes is the same.

[0133] Compared with traditional methods, the technical solution of the present invention has a simple identification method and higher identification efficiency. The traditional relational database identification method requires at least dozens of lines of code. Thanks to the cleverness of data modeling, the technical solution of the present invention does not require complex queries and can be implemented within ten lines of code, and the identification content is easy to visualize and analyze and deconstruct.

[0134] Use visualization technology to visualize the identification results of cyclic insurance behavior, deconstruct risk behavior through graphical presentation, intuitively display all parties involved in risk behavior, and efficiently implement risk warning and prevention.

[0135] This solution uses "sub-graph recognition" to identify "cyclic insurance" behavior in a dedicated business graph. The identified risk information is rich and naturally suitable for visual deconstruction, and can identify various behavioral patterns.

[0136] For the scenario of identifying cyclic insurance behavior, this solution can also identify multiple cyclic insurance behavior patterns, such as Fig. 9 As shown in Figure 1, after the customer takes out insurance, he borrows money against a policy and takes out insurance again. Fig.10 As shown, after the customer took out insurance, he borrowed money for multiple policies and insured them again, and the borrowing was completed in multiple times. Fig.10 The risk level of the circular insurance behavior shown is higher than Fig. 9 The risk level of the recurring insurance behavior shown.

[0137] That is, the more policy nodes corresponding to the same customer node included in each subgraph involved in the second risk business behavior, or the more policy nodes that have a second edge with the same customer node included in each subgraph involved in the second risk business, the higher the risk level of the second risk business behavior.

[0138] It should be noted that if Figure 1 As shown, step S2 and step S3 can be executed separately or together. If only the first risky business behavior is identified, only step S2 can be executed, that is, the complete execution order is S1-S2; if only the second risky business behavior is identified, only step S3 can be executed, that is, the complete execution order is S1-S3; if it is necessary to identify the first risky business behavior and the second risky business behavior, then steps S2 and S3 can be executed, and the execution order of steps S3 and S3 is not required, that is, the complete execution order is S1-S2-S3, or S1-S3-S2.

[0139] It should also be noted that the risk business behavior identification business graph in the technical solution of the present invention can support regular or irregular updates, and the update method can be automatic update or considered update, and the present invention does not limit this.

[0140] like Fig.11 As shown, the technical solution of the present invention also provides a method for identifying risky business behaviors of a life insurance company, further comprising:

[0141] S4, marking the customer nodes, policy nodes, sales agent nodes included in each subgraph involved in the first risk business behavior and the edges between the customer nodes, policy nodes, and sales agent nodes involved in the first risk business; and / or,

[0142] S5, marking the customer nodes and policy nodes included in each subgraph involved in the second risk business behavior and the edges between the customer nodes and policy nodes involved in the second risk business.

[0143] It should be noted that if Fig.11As shown, step S4 and step S5 can be executed separately or together. If only the first risky business behavior is identified, only step S4 can be executed, that is, the complete execution order is S1-S2-S4; if only the second risky business behavior is identified, only step S5 can be executed, that is, the complete execution order is S1-S3-S5; if it is necessary to identify the first risky business behavior and the second risky business behavior, steps S4 and S5 can be executed, and the execution order of steps S4 and S5 is not required, that is, the complete execution order is S1-S2-S3-S4-S5, or, S1-S2-S3-S5-S4, or, S1-S3-S2-S4-S5, or, S1-S3-S2-S5-S4, or, S1-S2-S4-S3-S5, or, S1-S3-S5-S2-S4.

[0144] Compared with traditional methods, the technical solution of the present invention has a simple identification method and higher identification efficiency. The traditional relational database identification method requires at least dozens of lines of code. Thanks to the cleverness of data modeling, the technical solution of the present invention does not require complex queries and can be implemented within ten lines of code, and the identification content is easy to visualize and analyze and deconstruct.

[0145] Use visualization technology to visualize the results of order splitting behavior identification, deconstruct risky behaviors through graphical presentation, intuitively display all parties involved in risky behaviors, and efficiently implement risk warnings and prevention and control.

[0146] The technical solution of the present invention identifies the "splitting orders" and "circular insurance" behaviors in a dedicated business graph through the "sub-graph recognition" method. The identified risk information is rich, naturally suitable for visual deconstruction, and can identify various behavior patterns.

[0147] In the technical solution of the present invention, a business graph for identifying risky business behaviors is established with insurance policies, customers, and sales agents as nodes, and with insurance relationships, sales relationships, and loan relationships among insurance policies, customers, and sales agents as edges; a first risky business behavior is identified based on the insurance policy nodes, sales agent nodes, and customer nodes included in the subgraph of the risky business behavior identification business graph, and the edges between the insurance policy nodes, sales agent nodes, and customer nodes; a second risky business behavior is identified based on the insurance policy nodes, customer nodes, and the edges between the insurance policy nodes and customer nodes included in the subgraph of the risky business behavior identification business graph, thereby effectively solving the problem of low efficiency in identifying order splitting behaviors and / or circular insurance behaviors caused by the prior art, and effectively improving the efficiency in identifying order splitting behaviors and / or circular insurance behaviors.

[0148] The technical solution of the present invention abandons the traditional data modeling method based on relational databases, and constructs a business graph dedicated to the identification of "order splitting" and "cyclic insurance" behaviors in the graph database based on the "graph model". Through clever data modeling and combining the advantages of graph databases in graph data analysis and mining, the "sub-graph recognition" method is used to efficiently identify the behaviors of order splitting and cyclic insurance, which reduces the difficulty of identification, improves the identification efficiency, and improves the accuracy and comprehensiveness of the identification results.

[0149] In the technical solution of the present invention, the customer nodes, policy nodes, sales agent nodes included in each subgraph involved in the first risk business and the edges between the customer nodes, policy nodes and sales agent nodes involved in the first risk business are marked; the customer nodes, policy nodes included in each subgraph involved in the second risk business and the edges between the customer nodes and policy nodes involved in the second risk business are marked, and the identification return result is still graph data (subgraph), which is easy to visualize and can fully deconstruct the various parties involved in the "order splitting" and circular insurance behavior.

[0150] In the technical solution of the present invention, the more sales agents included in each subgraph involved in the first risk business behavior, or the more policy nodes corresponding to each sales agent in each subgraph involved in the first risk business, the higher the risk level of the first risk business behavior corresponding to the subgraph; the more policy nodes corresponding to the same customer node included in each subgraph involved in the second risk business behavior, or the more policy nodes that have a second edge with the same customer node included in each subgraph involved in the second risk business, the higher the risk level of the second risk business behavior. It can quickly identify subgraphs with higher risks, reduce the company's losses and risks of fraud, and improve business quality. It can help life insurance companies efficiently identify "splitting orders" and "circular insurance" behaviors, fully "understand" the various representative behavior patterns involved in the "splitting orders" of sales agents and the "circular insurance behavior" of customers, and pay special attention to sales agents and customers who frequently engage in the above-mentioned risky business behaviors, provide risk reminders and warnings, guide business behaviors to be more standardized, and help the company develop high-quality development.

[0151] Embodiment 2

[0152] like Fig.12 As shown, the technical solution of the present invention also provides a system for identifying risk business behaviors of life insurance companies, including:

[0153] Establish module 101, take insurance policies, customers, and sales agents as nodes, and take insurance relationships, sales relationships, and loan relationships among insurance policies, customers, and sales agents as edges, to establish a risk business behavior identification business graph; wherein the risk business includes a first risk business and / or a second risk business, the first risk business is a bill splitting behavior, and the second risk business is a circular insurance behavior;

[0154] The first identification module 102 identifies the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in the subgraph of the business graph according to the risk business behavior, and identifies the first risk business; and / or

[0155] The second identification module 103 identifies the policy nodes, customer nodes, and edges between the policy nodes and customer nodes included in the subgraph in the business graph according to the risk business behavior, and identifies the second risk business.

[0156] The implementation process of the monitoring and establishing module 101, the first identification module 102, and the second identification module 103 in the second embodiment corresponds to the method steps in the first embodiment, and will not be repeated in this second embodiment.

[0157] In the technical solution of the present invention, a business graph for identifying risky business behaviors is established with insurance policies, customers, and sales agents as nodes, and with insurance relationships, sales relationships, and loan relationships among insurance policies, customers, and sales agents as edges; a first risky business behavior is identified based on the insurance policy nodes, sales agent nodes, and customer nodes included in the subgraph of the risky business behavior identification business graph, and the edges between the insurance policy nodes, sales agent nodes, and customer nodes; a second risky business behavior is identified based on the insurance policy nodes, customer nodes, and the edges between the insurance policy nodes and customer nodes included in the subgraph of the risky business behavior identification business graph, thereby effectively solving the problem of low efficiency in identifying order splitting behaviors and / or circular insurance behaviors caused by the prior art, and effectively improving the efficiency in identifying order splitting behaviors and / or circular insurance behaviors.

[0158] The technical solution of the present invention abandons the traditional data modeling method based on relational databases, and constructs a business graph dedicated to the identification of "order splitting" and "cyclic insurance" behaviors in the graph database based on the "graph model". Through clever data modeling and combining the advantages of graph databases in graph data analysis and mining, the "sub-graph recognition" method is used to efficiently identify the behaviors of order splitting and cyclic insurance, which reduces the difficulty of identification, improves the identification efficiency, and improves the accuracy and comprehensiveness of the identification results.

[0159] In the technical solution of the present invention, the customer nodes, policy nodes, sales agent nodes included in each subgraph involved in the first risk business and the edges between the customer nodes, policy nodes and sales agent nodes involved in the first risk business are marked; the customer nodes, policy nodes included in each subgraph involved in the second risk business and the edges between the customer nodes and policy nodes involved in the second risk business are marked, and the identification return result is still graph data (subgraph), which is easy to visualize and can fully deconstruct the various parties involved in the "order splitting" and circular insurance behavior.

[0160] In the technical solution of the present invention, the more sales agents included in each subgraph involved in the first risk business behavior, or the more policy nodes corresponding to each sales agent in each subgraph involved in the first risk business, the higher the risk level of the first risk business behavior corresponding to the subgraph; the more policy nodes corresponding to the same customer node included in each subgraph involved in the second risk business behavior, or the more policy nodes that have a second edge with the same customer node included in each subgraph involved in the second risk business, the higher the risk level of the second risk business behavior. It can quickly identify subgraphs with higher risks, reduce the company's losses and risks of fraud, and improve business quality. It can help life insurance companies efficiently identify "splitting orders" and "circular insurance" behaviors, fully "understand" the various representative behavior patterns involved in the "splitting orders" of sales agents and the "circular insurance behavior" of customers, and pay special attention to sales agents and customers who frequently engage in the above-mentioned risky business behaviors, provide risk reminders and warnings, guide business behaviors to be more standardized, and help the company develop high-quality development.

[0161] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for identifying risky business behavior of a life insurance company, characterized in that: include: A business graph for identifying risky business behaviors is established with insurance policies, customers, and sales agents as nodes, and with the insurance relationships, sales relationships, and loan relationships among insurance policies, customers, and sales agents as edges; wherein risky business behaviors include first risky business behaviors and / or second risky business behaviors, wherein the first risky business behaviors are splitting behaviors, and the second risky business behaviors are cyclic insurance behaviors; Identify the first risky business behavior by identifying the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in the subgraph of the risky business behavior identification business graph; and / or, According to the risk business behavior, the policy nodes and customer nodes included in the subgraph in the business graph and the edges between the policy nodes and the customer nodes are identified to identify the second risk business behavior.

2. A method for identifying risky business behavior of a life insurance company according to claim 1, characterized in that: In the risk business behavior identification business graph, each customer node represents a customer; each policy node represents an insurance policy; and each sales agent node represents a sales agent. There is a first edge and a second edge between the customer node and the policy node, and there is a third edge between the policy node and the sales agent node; the first edge is from the customer node to the policy node, representing an insurance relationship; the second edge is from the customer node to the policy node, representing a policy loan relationship; the third edge is from the sales agent node to the policy node, representing a policy sales relationship; the policy node includes policy type attribute information.

3. A method for identifying risky business behavior of a life insurance company according to claim 2, characterized in that: The first side is marked as an insurance type, the second side is marked as a loan type, and the third side is marked as a sales type; The first edge includes first attribute information and second attribute information, the first attribute information is the insurance time, and the second attribute information is the insurance effective time; The second side includes fourth attribute information and fifth attribute information, the fourth attribute information is the policy loan time, and the fifth attribute information is the repayment time of the policy loan.

4. A method for identifying risky business behavior of a life insurance company according to claim 3, characterized in that: According to the risk business behavior, the subgraph in the business graph includes the policy node, the sales agent node, the customer node, and the edges between the policy node, the sales agent node, and the customer node. The identification of the first risk business behavior specifically includes: If the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in a subgraph in the risk business behavior identification business graph meet the preset first risk identification rules, then the insurance behavior involved in the subgraph constitutes a first risk business behavior.

5. A method for identifying risky business behavior of a life insurance company according to claim 4, characterized in that: The first risk identification rule includes: The same customer node is connected to at least two policy nodes; and, The first type of policy node is connected to at least two different sales agent nodes, wherein the first type of policy node is at least two policy nodes connected to the same customer node; and, The first attribute information or the second attribute information of the first edge between the same customer node and the first type of policy node is in the same year and month; and, The policy type attribute information of the first type of policy nodes are all the same.

6. A method for identifying risky business behavior of a life insurance company according to claim 3, characterized in that: According to the risk business behavior identification business graph, the subgraph includes the policy node, the customer node, and the edge between the policy node and the customer node. The identification of the second risk business behavior specifically includes: If the policy nodes, customer nodes, and edges between policy nodes and customer nodes in the risk business behavior identification business graph meet the preset second risk identification rules, then the insurance behavior involved in the subgraph constitutes a second risk business behavior.

7. A method for identifying risky business behavior of a life insurance company according to claim 6, characterized in that: The second risk identification rule includes: The same customer node corresponds to at least two policy nodes, and there is a second edge between at least one of the first-type policy nodes and the customer node, wherein the first-type policy nodes are at least two policy nodes corresponding to and connected to the same customer node; and, The first attribute information of the first edge of the first category is less than the fourth attribute information of the second edge of the first category, wherein the first edge of the first category is the first edge between the same customer node and the policy node having a second edge with the customer node; the second edge of the first category is the second edge between the same customer node and the policy node having a second edge with the customer node; and, The fourth attribute information of the first-category second edge is less than the first attribute information of the second-category first edge, wherein the second-category first edge is the first edge between the same customer node and the policy node between which no second edge exists; and, The fifth attribute information of the second edge of the first category is greater than the first attribute information of the first edge of the second category; and, The policy type attribute information of the first type of policy nodes are all the same.

8. A method for identifying risky business behavior of a life insurance company according to claim 1, characterized in that: The more sales agents included in each subgraph involved in the first risk business behavior, or the more policy nodes corresponding to each sales agent in each subgraph involved in the first risk business, the higher the risk level of the first risk business behavior corresponding to the subgraph; The more policy nodes corresponding to the same customer node included in each subgraph involved in the second risk business behavior, or the more policy nodes that have a second edge with the same customer node included in each subgraph involved in the second risk business, the higher the risk level of the second risk business behavior.

9. A method for identifying risky business behavior of a life insurance company according to any one of claims 1 to 8, characterized in that: include: Marking the customer nodes, policy nodes, sales agent nodes included in each subgraph involved in the first risk business behavior and the edges between the customer nodes, policy nodes, and sales agent nodes involved in the first risk business; and / or, Mark the customer nodes and policy nodes included in each subgraph involved in the second risk business behavior and the edges between the customer nodes and policy nodes involved in the second risk business.

10. A system for identifying risky business behavior of a life insurance company, characterized by comprising: A module is established, with insurance policies, customers, and sales agents as nodes, and the insurance relationship, sales relationship, and loan relationship between insurance policies, customers, and sales agents as edges, to establish a risk business behavior identification business graph; wherein the risk business includes the first risk business and / or the second risk business, the first risk business is the order splitting behavior, and the second risk business is the circular insurance behavior; A first identification module identifies the policy nodes, sales agent nodes, customer nodes, and edges between the policy nodes, sales agent nodes, and customer nodes included in the subgraph of the business graph according to the risk business behavior, and identifies the first risk business; and / or The second identification module identifies the policy nodes, customer nodes, and edges between the policy nodes and customer nodes included in the subgraph in the business graph based on the risk business behavior, and identifies the second risk business.

Images