<![CDATA[The telecommunications industry faces intense competition that makes customer churn a critical threat to business sustainability. This study proposes an integrated analytical framework that combines churn prediction, customer segmentation, and model interpretability to support data-driven retention strategies. Three machine learning algorithms Logistic Regression, Random Forest, and XGBoost were trained on the IBM Telco dataset (n = 7,043) with class imbalance handled through SMOTE. K-Means clustering was then applied to predicted-churn customers, and SHAP (SHapley Additive exPlanations) was used to interpret the best-performing model both globally and at the segment level. Random Forest achieved the best performance (Accuracy 0.7763, F1-Score 0.6114, ROC-AUC 0.833), and three customer segments were identified: New High-Risk, Loyal High-Value, and Low-Tier Economic. The per-segment SHAP analysis revealed heterogeneous churn drivers across segments, enabling the formulation of more targeted and actionable retention strategies. The integration of churn prediction, segmentation, and SHAP therefore provides a more precise foundation for data-driven retention decision-making in the telecommunications industry.]]>
Copyrights © 2026
