Insurance anti-gang fraud method based on graph rule engine

Abstract

The invention discloses an insurance anti-gang fraud method based on a graph rule engine, and the method comprises the steps: employing the existing historical data of an insurance enterprise, and building a graph for various types of data relationships between people and vehicles, between people and cases, between vehicles and cases, between vehicles and enterprises, and between cases and accounts; a, the relationship between the person and the vehicle comprises a vehicle owner, a driver, an insured person and an insured person; b, the incidence relation between the person and the case comprises the wounded person; c, the relationship between the vehicle and the case comprises an accident vehicle and three vehicles; d, the relationship between the enterprise and the vehicle is that the enterprise is a vehicle owner; e, the relation between the case and the account is that the account is the claim settlement account of the case, so that a certain payment amount is generated; the mapping process is based on the total related data of an enterprise, and a large graph containing ten millions and even hundred millions of nodes is generated accordingly. According to the method, the rule set and the scoring sequence of the graph rule engine can be flexibly adjusted, the graph splitting process can be carried out in a nested mode, the process is independent of service logic, and program design and implementation do not need to be frequently changed.

Description

technical field [0001] The invention relates to the technical field of the insurance industry, in particular to an insurance anti-gang fraud method based on a graph rule engine. Background technique [0002] In the insurance industry, gang-like fraud is one of the most difficult diseases to detect and prevent. Traditional anti-fraud methods are based on a known set of rules, based on the relationship between people and people based on relevant data such as insurance and claims, and then perform manual analysis and identification. However, there are three main shortcomings of the traditional method: [0003] 1. The set of rules used in calculating the relationship between people must be based on past experience to determine meaningful rules, and the weight of the rules (if any) must also be determined in advance. These historical experiences may not fully reflect the current actual situation, resulting in inaccurate calculation results of association relationships. [0004...

AI Technical Summary

Problems solved by technology

These historical experiences may not fully reflect the current actual situation, resulting in inaccurate calculation results of the relationship

[0004] 2. If the rule set is adjusted (such as adding or reducing rules, and the weight of some rules changes), then the program must be rewritten and run to recalculate the relationship between people. Environment costs can be high

[0005] 3. It is impossible to use the full amount of data to discover new relevant rules and automatically adjust them

Images