Combining predictive accuracy and interpretability: a data-driven approach to telecom churn analysis.

Area of Science:

• Machine Learning
• Artificial Intelligence
• Data Science

Background:

• Customer acquisition is costly, making customer retention vital for telecom profitability.
• Predicting customer churn is a significant challenge in competitive telecommunications markets.

Purpose of the Study:

• To introduce XCL-Churn, an explainable ensemble learning framework for robust and interpretable customer churn prediction.
• To integrate XGBoost, CatBoost, and LightGBM using a soft-voting meta-architecture for enhanced churn prediction.

Main Methods:

• Employed a data preprocessing pipeline including iterative Bayesian Ridge imputation, multi-stage scaling, and hybrid Boruta-Random Forest feature selection.
• Addressed class imbalance using Synthetic Minority Oversampling Technique (SMOTE).
• Integrated XGBoost, CatBoost, and LightGBM models within a soft-voting ensemble and applied Explainable AI (XAI) techniques (LIME, SHAP).

Main Results:

• The XCL-Churn ensemble achieved high performance metrics: 97.44% accuracy, 93.82% precision, 87.82% recall, and 91.25% F1-score.
• Demonstrated superior predictive performance and computational efficiency compared to traditional methods.
• XAI techniques provided transparency into key behavioral and financial drivers of customer churn.

Conclusions:

• XCL-Churn offers a robust, interpretable, and computationally efficient solution for customer churn prediction in the telecom industry.
• The framework's ability to identify key churn indicators enhances strategic decision-making for customer retention.