Customer Churn Prediction in Telecommunications Using Machine Learning Models

Customer churn prediction has a great impact on the long-term profitability and competitiveness of telecommunications companies. This study uses telecommunications customer data, employs machine learning methods to build a customer churn prediction model, and identifies key influencing factors. By using majority class under-sampling and threshold optimization to alleviate the problem of class imbalance, the overall model performance is improved. Experimental results show that logistic regression, radial basis function kernel support vector machine, random forest, and gradient boosting all have strong predictive ability. The random forest model performs the most evenly, with a recall rate of 0.84 and an F1 score of 0.79 and is selected as the optimal model. The feature importance analysis shows that service availability, Internet service type and cumulative billing amount are the core factors affecting customer churn. These findings provide effective forecasting models and insights for the telecommunications industry to improve customer retention strategies.