Churn prediction for internet service provider

Churn prediction is a customer relationship process that specifies predictions for customers who are inclined to churn in future through modelling customer behavior patterns. It costs more to acquire a customer than to retain a customer. In this sense, the predictions which are made with churn prediction support promotion activities executed to avoid subscription cancellation of existing customers. Nowadays, telecommunication companies maintain churn prediction with various applications as a systematic process. Also this thesis is written on the basis of customer data and behavior analysis of an internet service provider operating in telecommunication sector. Within the knowledge discovery process framework, explored as a result of realized literature survey, two phased solution model is created with the help of data mining applications. Developed two phased solution model is designed to run effectively with time series clustering and classification algorithms. Performance indicators are evaluated comparatively with respect to k-means, hierarchical clustering algorithms for time series clustering and support vector machines, recursive partitioning for classification algorithms.