Empirical Analysis of Multilayer Perceptron and Different Data Mining Techniques for Classifying Customer Data for Effective Customer Churn Prediction

In this competitive world, mobile communications market lean towards to reach a saturation state and faces an aggressive struggle. This state forces the telecom businesses to center their care on keeping the clients together instead of constructing a big customer base. This causes to mine earlier data containing unidentified telecom user segments. This data needs applying different procedures agreeing to the arrangement of the data sets to be analysed. Telecommunication marketplace is quickly emerging and becoming viable in several nations due to protocols, new computer and communication expertise. In this state of affairs, data mining is mandatory for considering the business requirements, describing the communications model, consuming sources efficiently and refining facility excellence. One of the main concerns of telecommunications companies is the subscriber retention. This sector has an annual churn rate of approximately 30% and is on top of the list. Since it is more costly to acquire a new subscriber than to retain an existing one, it is important for these telecommunications operators to identify subscribers that are at risk of churning by using predictive models. We have presented a classifier named Multi Layer Perceptron with PCA. In this paper, we present analysis (accuracy, precision, recall, F-measure) that predicts churn in the pre-paid mobile telecommunications industry on various datasets that contains call details records.