Telecom Customer Churn Forecasting Using Machine Learning: A Data-Driven Predictive Framework

Abstract

Customer churn is a significant challenge for businesses, impacting both short-term profits and long-term sustainability. Accurately predicting churn is essential for companies focused on retaining valuable customers and reducing acquisition costs. This paper explores the development and evaluation of a Customer Churn Prediction Model using several Machine Learning (ML) algorithms, such as Logistic Regression, Random Forest, XGBoost, Decision Tree, K-Nearest Neighbors (KNN), and Deep Learning, implemented on the RapidMiner platform. The analysis uses a publicly available telecommunications dataset from Kaggle, containing customer demographics, service usage, and billing information. The study follows key stages in the data science process, including data preparation, feature engineering, model training, and evaluation. Model performance is measured using metrics like Relative Mean Squared Error, Absolute Error, and Correct Predictions. While Deep Learning achieved the highest accuracy, Logistic Regression was the most interpretable and reliable. The findings highlight the importance of AI/ML in churn prediction, helping businesses optimize strategies and improve customer retention.