Retaining Customers Using Clustering and Association Rules in Insurance Industry: A Case Study

Abstract

This study clusters customers and finds the characteristics of different groups in a life insurance company in order to find a way for prediction of customer behavior based on payment. The approach is to use clustering and association rules based on CRISP-DM methodology in data mining. The researcher could classify customers of each policy in three different clusters, using association rules. At the end of study the characteristics are defined and given to the company, so they could implement CRM strategies based on the newly found differences. Attention to the income and cash earning comes before paying attention to other problems. In most of the companies in developing countries, infrastructural problems of the company like earning enough income prevent the company from effective research implementation on advanced strategies. So this study focuses on basic problems. Utilizing data mining approach to classify customers in life insurance is a new approach among insurance companies in Iran. There are some research in relation to the CRM and data mining, but the contribution of this study is to investigate two new attributes plus those common attributes used before in studying customer behavior; the two attributes are "payment type" and the "purchaser". In order to have a framework, all the process is embedded in CRISP-DM methodology.