Pastoralist Optimization Algorithm Approach For Improved Customer Churn Prediction in the Telecom Industry

Abstract

In recent times, Telecom Industry customer churn has been a serious problem making it difficult to survive in the fierce competition of the industry. Survival in the industry and retention of the existing customers has become very important. Practitioners and academicians are now faced with the challenge of getting to predict likely customer churn, through predictive modeling techniques to predict potential customers who are likely to churn. Customer churn affects the revenue of the company because it cost more to acquire new customer than retaining old ones. When a company allocates its dedicated resources to retain these customers, it greatly controls the rate at which dissatisfied customers leave the company. Several techniques have been studied and we present an overview of resent works on churn prediction .Our work uses Artificial Neural Network approach for prediction of customers intending to switch over to other operators. This study uses Pastoralist Optimization Algorithm (POA) to enhance the Artificial Neural Network (ANN) by working on multiple attributes from Telecom Company’s dataset with sample results. The results obtained showed that the proposed POA algorithm selected fewer attributes of ten out of fifteen for the telecom churn prediction and had a prediction accuracy of 97.0% compared to the ordinary unenhanced ANN which used the entire 15 attributes but had a prediction accuracy of 93.6%.