Customer Churn Prediction Using Machine Learning: A Case Study of Libyan Internet Service Provider Company

Artificial intelligence (AI) is a field of science and technology that simulates human intelligence to develop intelligent machines. Machine Learning (ML), on the other hand, is a subset of AI that enables computers to learn from data without being explicitly programmed. In the area of business, data science is the practical application of ML, AI, and statistics, which focuses on extracting valuable insights from data. Nowadays, data science relies heavily on ML techniques, which are used for a variety of applications. One of its applications is Customer Churn Prediction (CCP). Customer churn (or attrition) is the tendency of customers to stop using a company's products or services. It is important to predict churn where businesses can proactively target the customers, who are at high risk of leaving and create marketing campaigns to retain them. This study aims to build a predictive model to predict customer churn for Internet Service Providers (ISPs). Several experiments have been conducted on a dataset obtained from a Libyan ISP company to train the models. The dataset distribution was imbalanced, with only 5% of customers churned. To mitigate this, various oversampling techniques were used. In the proposed system, five diverse classifiers were used. Namely, Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Gradient Boosting Machine (GBM) tree algorithm, and Extreme Gradient Boosting (XGBoost) algorithms due to their relevance in this predictive context. The experimental findings showed that the XGBoost tree model yielded the most favorable outcomes, achieving an AUC value of 94%.