Enhancing customer retention in telecom industry with machine learning driven churn prediction

Area of Science:

• Machine Learning
• Data Science
• Business Analytics

Background:

• Customer churn poses a significant business challenge, necessitating effective prediction for retention strategies.
• Imbalanced and diverse customer data distributions complicate accurate churn prediction models.
• Existing literature highlights the need for advanced data balancing techniques in churn analysis.

Purpose of the Study:

• To introduce and evaluate a novel Ratio-based data balancing technique for improving churn prediction accuracy.
• To compare the effectiveness of the proposed technique against traditional data resampling methods.
• To assess the performance of various machine learning algorithms, including ensemble methods, on balanced datasets.

Main Methods:

• Development of a novel Ratio-based data balancing technique to address data skewness.
• Evaluation of machine learning algorithms: Perceptron, Multi-Layer Perceptron, Naive Bayes, Logistic Regression, K-Nearest Neighbour, Decision Tree, Gradient Boosting, and Extreme Gradient Boosting (XGBoost).
• Comparison of the proposed Ratio-based technique with traditional Over-Sampling and Under-Sampling methods using metrics like Accuracy, Precision, Recall, and F-Score.

Main Results:

• The Ratio-based data balancing technique demonstrated superior performance over traditional methods in churn prediction.
• Ensemble algorithms, specifically Gradient Boosting and XGBoost, outperformed single machine learning models.
• The XGBoost method with a 75:25 ratio yielded the most promising results for churn prediction accuracy.

Conclusions:

• The Ratio-based data balancing technique is effective in enhancing churn prediction accuracy by addressing data imbalance.
• Ensemble machine learning methods, particularly XGBoost, are highly effective for customer churn prediction when applied to balanced datasets.
• Optimized data balancing and ensemble modeling offer significant improvements for customer retention strategies.