Application Research Of Data Mining On Insurance Customer Subsection

During long term of informationlize, insurance industry has been one of themost .How to make use of the amount of data to support managers to make decisionbecomes the next direction of insurance's information development. Customersubsection is one of the basic tasks in insurances or sale industry. Traditionalsubsection method used in insurance field is limited facing the agent problem. In thatcase, setting up consummate customer subsection solution is crucial.Data mining technology is a new cross filed of machine learning, statistics anddatabase technology. It is the collection of engineering methods proposed for solvinglarge practical problems of machine learning, pattern recognition, database technologyetc. The purpose of data mining is to discover the implicit knowledge or rule, and tohelp human expert to make decision. Neural network by right of its high ability tosolve nonlinear problem and outlier becomes the representative of classify methods.This thesis using cluster and neural network makes attempts to solving customersubsection for one insurance company. The main contributions are following:Firstly, setting up insurance customers subsection flow basing on CRISP's flow forData Mining.Secondly, setting up data structure for this application according to businessrequirement. Use data preparation technology to conduct the data, which suppliessupport to algorithm's efficiency.Thirdly, analyzing the activation functions and topology of neural network, comparingthe results of MultiLayerPerceptron and RBF neural.Last, constructing B/S customer management system using JAVA. This systemimplements the function of data structure setting up, data preparation, clusteringdynamically and classification.