Predicting customer churn in telecom companies through a machine-learning approach

Abstract

The research inspects the application of machine learning approaches to forecasting customer attrition in telecommunication companies. Machine learning models such as extreme gradient boosting, random forest, k-nearest neighbour, adaptive boosting, support vector machine, and logistic regression were used to forecast and compared the best model and analysed churn behaviour. Cross-validation techniques were applied to enhance model performance, revealing critical predictors of churn such as contract length, customer tenure, and service usage patterns. The results emphasised the effectiveness of machine learning in accurately identifying potential churners. Furthermore, the study emphasises the importance of leveraging predictive analytics to proactively address customer attrition, enabling telecommunication companies to devise targeted retention strategies and enhance customer satisfaction and loyalty.