BaY cP: A novel Bayesian customer Churn prediction scheme for Telecom sector

Abstract

The current Telecom sector is highly competitive due to increased Mobile Number Portability (MNP) of users. The ease of MNP and plenty of switching options between Telecom providers, leads to rise in attrition, known as the churn behavior in customers. Customer is always in pursuit of better services at cheaper rates from service vendors. Thus, in this competitive Telecom market, the providers face a dual issue to retain loyal customers, as well as attract new potential customers by providing cheap data plans and free calling options. Thus, this unreasonable demand vs. supply rate to satisfy such customers effects the profitability of the company, which is a serious concern. Thus, to mitigate such fluctuations, termed as customer churn (CC) behavior, the paper a novel scheme BaYcP, that addresses the CC problem in two phases. In the first phase, based on customer data-sets, risk profiling score (RPS) is generated based on descision trees, and is compared to a threshold value. Then based on scores higher than threshold, an optimal prediction model is built based on bayesian classifier on appropriate selected features. The model is trained and validated to achieve and accuracy of 97.89% which outperforms other state-of-the art approaches.