我国险企运营中道德风险甄别问题研究——以大数据Hadoop聚类分析技术为视角  

A Research on Chinese Insurers' Moral Hazard Screening in Operation:From the Big Data Hadoop Clustering Analysis Technology Perspective

作　　者：王海巍[1] 

机构地区：[1]东北财经大学金融学院,辽宁大连116025

出　　处：《保险研究》2016年第2期59-67,共9页Insurance Studies

基　　金：中国保险监督管理委员会大连监管局重点研究课题"大连市海水养殖风险评价与巨灾保险运行机制研究(2015)"资助

摘　　要：2014年国务院发布保险"新国十条",明确提出推进保险业改革开放,鼓励保险产品服务创新。本文将契约理论与保险精算思想结合,研究保险实务中道德风险甄别问题。基于险企运营动态数据库中的业务数据流,对涉及投保、承保、核保、案件受理、勘查、核赔、理赔等环节关键数据字段聚类分析筛选,通过建立数理模型观测、估计道德风险阈值,提出风险预警。研究发现因传统道德风险判别方法主要基于静态样本数据进行经验认定,存在负相关性和滞后性。建议保险公司基于动态Hadoop模型进行风险因子聚类分析,预判高危主体,抑制道德风险。同时从行业机制建设、技术应用角度提出建议。In 2014,the State Council issued the ＂Ten New Provisions＂ for insurance industry＇ s development, explicitly promoting reform and opening up of the industry and innovating on insurance products and services. This article combined the contract theory and actuarial ideology for the study of moral hazard screening in insurance operation. Based on insurance firms＇ operating data stream in the dynamic database, it conducted aclustering analysis screening on insurance application, underwriting, risk selection, case acceptance, investigation, claims ratification, claims settlement and other key aspects of the data segment, established a mathematical model for observing and es- timating the moral risk threshold, and then offered warnings on related risks. The study showed that due to the fact that traditional moral hazard identification method was mainly based on experience rating of static sample data, there was a negative correlation and time lag. It suggested that insurance companies should adopt the dynamic risk factor model based on Hadoop clustering analysis to predict high-risk objects and restraint moral hazards. It also put forth proposals on industry mechanism construction and technology application.

关 键 词：险企运营 道德风险 大数据 聚类分析 

分 类 号：F840.4[经济管理—保险]