An efficient hybrid clustering to predict the risk of customer churn

Companies are facing challenges of revenue losses due to increased market competition hence issue of the loss of customers. Many are urged to identify the reasons of losing customers by measuring customer loyalty thereby to take appropriate steps to regain the lost customers and hence increase their revenue. The movement of customers from the current company to another company or competitor is termed as customer churn. Customer churns can be minimized by analyzing the past history of the potential customers of a company systematically. In addition, decision makers are always faced with imprecise operation management problem. Hence the need for prediction mechanism for churn management and continuous updation of appropriate strategies has become more important today's competitive world. To deal with these issues a hybrid method composed of Axiomatic Fuzzy Set (AFS) and parallel DBSCAN clustering is considered. The experimental results indicate that Parallel DBSCAN takes much less running speed then parallel K-means clustering.