Research On Data Mining Modle For Car Insurance Personalized Pricing System

Nowadays, since the pricing ability of the car insurance has a great influence on thebusiness capacity of insurance companies, car insurance pricing becomes a hot topic in theresearch of insurance industry. Most current researches about car insurance pricing arebased on linear or generalized linear model which forms the basic rate table to use in theprocess of car insurance pricing. However, these methods can’t meet the requirements ofthe developing insurance industry.This paper mostly introduced the data mining algorithm and model into car insurancepricing instead of the linear or generalized linear models. What’s more, research whetherthe modeling and verification of the data mining model can be integrated in the pricingsystem.In order to establishing a personalized car insurance pricing system which is based ondata mining model and practical for insurance companies, this paper studies from thefollowing four aspects.Firstly, we introduce the related technologies of the system including the overalllogical framework, ETL framework, data storage model and data mining model integrationframework.Secondly, we carry on the research of modeling approach of data mining techniques,and choose the extreme learning machine (ELM) algorithm for data modeling.Thirdly, from the view of business understanding, we study the business model ofpersonalized car insurance pricing and the target factor and controlling factors of businessmodel. Because car insurance pricing involves the expertise of many insurance industries,it is necessary to have in-depth knowledge of car insurance pricing business to establish apractical pricing model of car insurance. And, at the same time, with exploring andunderstanding a small amount of data obtained from an insurance company, the papercleans and transforms the data. At the same times, the target factors and the control factorsare analyzed deeply to know whether they are useable in the model, and the final choice ismade. Finally, on the basis of the data, the system is constructed and the data are verified totest the availability of model and the feasibility of system. Therefore, the study is feasibleand valuable, which achieves the goal of the researching project.