Modeling Churn Behavior of Bank Customers Using Data Mining Techniques in MEHR Institute (SVM+Fuzzy Rules)

In recent years, we have been faced with increasing growing use of data mining and database and parallel to this, new science and technologies have been developed to make the best of these huge volumes of data in customer relationship management, optimized use of data is a competitive advance for companies and organization. And if so, they can deserve their old customers and acquire new ones. In this regard, the banking industry is one of those industries that deal with huge volume data of the customers. In this article, we have considered customers churn in Mehr finance and credit institution. In the first step of this thesis, the literature review results is reported and based on the review, different aspects of problem properties is investigated and then, the required data is gathered. The churn modeling of data is performed in a hybrid form: first, we have applied fuzzy modeling for classifying customers in 3 categories as Active, Moderate and weak. Then, we have made use of support vector machine for churn modeling. Numerical results of the problem model are compared with logical approach. Finally, based on these numerical results, we can conclude that the hybrid Fuzzy – SVM approach is a more exact and useful approach for churn prediction.