An extended PPFCM- ANN Model for Telecommunication Customer Retention

Abstract

The term ‘Churn’ indicates a situation when an existing customer shows no longer interest to continue his or her participation in utilizing the services of a company. Hence Churn Prediction plays a crucial dynamic role that paves the way for the sustainable growth of the organization importantly in the ever-challenging telecommunication industry. This article intends to propose a framework to forecast a customer by hybrid probabilistic possibilistic fuzzy C-means clustering (PPFCM) along with Ensemble classification (PPFCM-Ensemble) techniques. This paper involves two modules: (1) PPFCM clustering techniques used for clustering module (2) Ensemble classifiers used for customer forecasting component module. During the training process, the train dataset is assembled into groups, with the assistance of PPFCM clustering techniques. The acquired segmented data is utilized in the following ensemble classifier and this mixture model is utilized for forecasting churn prediction process. During the testing procedure, the segmented test information selects the most precise ensemble classifier which relates to the nearest group of the test information, as indicated by least distance. At last, to forecast the churn customer using the proposed hybrid PPFCM-Ensemble model. Three different experiments are performed, it is proved that the proposed Hybrid PPFCM-Ensemble Model affords maximum accuracy in comparison with any solitary models.