A new neural network based customer profiling methodology for churn prediction

Abstract

Increasing market saturation has led companies to try and identify those customers at highest risk of churning. The practice of customer churn prediction addresses this need. This paper details a novel approach and framework for customer churn prediction utilising a Neural Network (NN) approach. The methodology for customer churn prediction describes a predictive approach for the identification of customers who are most likely to churn in the future. This is a departure from current research into customer churn which tries to predict which customers are most likely to instantaneously churn. A real life case study from industry is presented here to illustrate this approach in practice. Future research will include the enhancement of this approach for more accurate modelling of collective systems. © 2010 Springer-Verlag Berlin Heidelberg.