Predicting customer churn in telecommunication service provider industry using Random Forest / Wan Muhammad Naqib Zafran Wan Roslan

This project addresses the challenge of customer churn in the Telecommunications Service Provider (TSP) industry by focusing on the Random Forest algorithm for predictive modeling. This study aims to throughly explore the Random Forest algorithm, develop a robust Random Forest customer churn predictive model, and evaluate its performance in predicting customer churn within the internet service provider sector. The specific objectives include studying the complexity of the Random Forest algorithm, constructing a model adjusted to accurately predict customer churn, and conducting thorough testing and evaluation of the model's accuracy. Through many experimentation, it was found that a model with 20 trees, a maximum depth of 5, and a maximum of 8 features yielded the highest accuracy at 79%, with an area under the curve of 0.79 for the Receiver Operating Characteristics. The outcomes of this research are poised to contribute significantly to the improvement of revenue, customer satisfaction, and provide valuable insights for data scientists and analysts engaged in similar predictive modeling endeavors within the TSP industry.