动态异质网络表征学习的车险理赔反欺诈识别研究  

Dynamic Heterogeneous Network Representation Learning for Fraud Detection in Auto Insurance

作　　者：潘怡君 梁变 张泷 那崇宁 PAN Yijun;LIANG Bian;ZHANG Long;NA Chongning(Financial Technology Research Center,Zhejiang Lab,Hangzhou 311121,China)

机构地区：[1]之江实验室金融科技研究中心,杭州311121

出　　处：《计算机工程与应用》2024年第24期322-330,共9页Computer Engineering and Applications

基　　金：国家重点研发计划项目(2022YFB4501400);之江实验室青年基金。

摘　　要：针对车险理赔数据呈现多源异构形态及历史数据量较大的问题,提出一种基于动态异质网络表征学习的车险理赔反欺诈识别方法。利用图结构将车险理赔案件中不同结构多种属性的节点统一表示成向量形式,再利用传统的机器学习方法实现欺诈识别。根据车险理赔案件的欺诈类型,设计五条相应的随机游走规则,从多个角度描述车险欺诈事件;提出一种动态异质网络节点选择方法,辨识与新采集车险案件相关的节点,并计算历史案件中出现频次较高的节点,在新的时间戳动态更新这部分节点的随机游走路径及节点的向量表示。利用真实的车险理赔数据验证所提出算法的有效性,分别从欺诈的识别率、误报率、准确率、模型运行时间、节点选择数量以及模型窗口大小方面实现。Since the challenges inspired by the diverse and heterogeneous of the data and the large amount of historical data,a dynamic heterogeneous network representation learning method for fraud detection in the auto insurance is pro-posed.The graph is utilized to represent different structure and rich attribute nodes as vectors,and traditional machine learning algorithm is employed for fraud detection.Firstly,five random walk rules are designed based on the fraud types in the auto insurance,which provide multiple perspectives to describe fraud events.Secondly,a dynamic heterogeneous network node selection method is proposed to identify nodes relevant to newly collected auto insurance cases and the fre-quency of node in historical cases is calculated.The random walk paths and vector representations of these nodes are dynamically updated at a new timestamp.Finally,the effectiveness of the proposed algorithm is tested using the real auto insurance data,considering the fraud detection rate,fraud alarm rate,accuracy of fraud detection,running time,number of nodes and window size.

关 键 词：动态性 异质网络 表征学习 随机游走 欺诈识别 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]