Regression-Based Machine Learning Framework for Customer Churn Prediction in Telecommunication Industry

Abstract

Customers’ movement from one telecom provider to the other has become a foremost issue in the telecommunication industry. This exacting issue has engendered stiff competition among vendors in the telecommunication industry to retain their customers. This competition is consequent upon the fact that it is more costly to acquire new customers than it takes to maintain the existing ones. The ability to make an accurate prognosis about customers who are likely to churn, and to offer incentives to retain them, places such telecom providers on a foundational platform to stand in the market. Recent studies in churn prediction utilized a single machine learning model that the results cannot be easily generalized to a new dataset or new scenario. In addition, these machine learning models are complex and with high computational time. In this study, we propose a comprehensive and computationally efficient Regression-Based Machine Learning Framework for Customer Churn Prediction in Telecommunication Industry. We evaluated nine different Regression Models and compare their performances. Moreover, we evaluated and determine which model is best suited to the proposed approach. The models were evaluated using four commonly used Regression based metrics such as Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), and Coefficient of Determination (R2). The best accuracy was recorded by Lasso Regression, with MAE, MSE, RMSE, and R2 of 7.77E−02, 1.21E−02, 1.10E−01, and 0.981407 (98%), respectively. This result shows that the Lasso Regression-based model performed better, realized the line of best fit, fit well with the observed data, and guarantee better predictions when deployed using the proposed approach.