基于数据挖掘的保险客户风险—贡献评级管理  

Risk-contribution Rating Management of Insurance Customers Based on Data Mining

作　　者：刘晓葳[1] 

机构地区：[1]厦门大学经济学院,厦门361005

出　　处：《保险研究》2013年第3期100-109,共10页Insurance Studies

摘　　要：保险公司的经营特性决定其成本与收益主要源于理赔支出与保费收入,也从而决定了保户风险、贡献分析是保险公司成本-收益管理的重要部分。数据挖掘方法从投保人意愿透露的特征信息入手,从风险与贡献的双重角度提取客户特征,实现客户细分,使企业得以通过评级管理等手段实现其风险控制及利润最大化要求。本文采用罗吉斯回归、CHAID模型、CART模型以及Aprior算法等方法对台湾某知名保险公司21年间客户特征、客户保单、理赔信息等资料进行综合挖掘,以获得具备风险及贡献指向性的客户特征变量,基于精确性与增益度进行模型比较,获取挖掘结果,构建客户风险-贡献特征矩阵,进而提出相应的保险公司管理策略。The operating characteristics of insurance companies determine that their costs and benefits mainly arise from claims payments and premiums, so the client risk-contribution analysis is the core of the cost and revenue management of them. Data mining methods can abstract both risk and contribution features of customers to build a riskcontribution matrix to achieve the aim of customer segmentation and meet the management needs of risk-minimizing and profits-maximizing. Using the logistic regression, the CHAID model, the CART model and the Aprior data mining algorithm on 21-year annual data of a well-known Taiwan insurance company,in aspects of customer profiling, customer policies and claims information, the paper intended to obtain customer characteristic variables that could reveal risk and contribution features. Then, based on consideration of accuracy and the gain degree of each model, the paper built up a creditable risk-contribution matrix to classify customers and to provide strategic advice.

关 键 词：数据挖掘 保险 风险-贡献特征矩阵 分类评级 

分 类 号：F222.3[经济管理—国民经济]