Comparative Evaluation of Machine Learning Models for Customer Churn Prediction in the Telecom Sector

The phenomenon of churn has been a critical problem that has persisted in the telecom industry and its effect on the profitability of the companies and the strategies used in trying to retain customers. The analysis of customer churn through the use of machine learning is increasing due to the effectiveness exhibited in customer identification that may leave. This paper provides a literature review of the different types of machine learning algorithm used in the context of churn prediction in the telecom industry. Algorithms such as, Logistic regression, Decision tree, Random forest, SVM and Neural networks are tested and assessed on a telecom dataset. The effectiveness of all models is determined by using a number of validation indices such as accuracy, precision, recall, F1 measure, and ROC curve. Data also show that some algorithms are better than others and that the best algorithms are Random Forest and the Support Vector Machine. Finally, insights regarding certain aspects of customer retention management and the possible reduction of organisational losses due to churn are derived from these findings. This study advances all the efforts in the application of machine learning for better customer churn prediction and is helpful for more practioners in the industry. In addition, the suggestions for the future work, such as the development of deep learning models and the real-time data application, are also provided.