Analyzing Customer Churn Prediction Using Machine Learning in Telecommunications

Thesis Abstract

The unprecedented growth and competitive nature of the telecommunications industry have amplified the importance of understanding and managing customer churn to sustain profitability and market share. Customer attrition poses significant financial challenges, with studies estimating that acquiring new customers can be five to twenty-five times more expensive than retaining existing ones. This study aims to develop an effective predictive model for customer churn using advanced machine learning techniques to assist telecommunications providers in proactive retention strategies. The specific objectives are to analyze the key demographic, usage, and service-related factors influencing customer churn; compare the performance of various machine learning algorithms, including Random Forest, Support Vector Machine, and Gradient Boosting; identify the most significant predictors of customer attrition; and formulate a practical framework for deploying churn prediction models in operational environments. The research adopts a quantitative, explanatory research design, employing a cross-sectional approach to analyze a comprehensive dataset obtained from a major telecommunications company operating in a metropolitan region. The population comprises active prepaid and postpaid customers, totaling approximately 250,000 subscribers. A stratified random sampling technique is applied to select a representative sample of 5,000 customers, ensuring proportional representation across different customer segments and tenure periods. Data collection instruments include structured transactional and service usage records, demographic surveys, and customer feedback data, all obtained through the company's Customer Relationship Management (CRM) system. Ethical clearance was secured prior to data collection, guaranteeing confidentiality, anonymity, and adherence to data protection protocols. Data preprocessing involves cleaning, normalization, and feature engineering to prepare the dataset for analysis. The study employs a multi-model machine learning approach, evaluating the predictive performance of algorithms such as Random Forest, Support Vector Machine (SVM), Gradient Boosting Machines (GBM), and Logistic Regression as a baseline. Model training and validation are conducted using k-fold cross-validation to prevent overfitting, with performance metrics including accuracy, precision, recall, F1-score, and the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). Feature importance analysis identifies the most influential variables, guided by the theory of customer satisfaction and switching behavior models, including the "Customer Dissatisfaction-Exit" theory and the "Push-Pull-Mooring" framework. Expected findings suggest that machine learning models, particularly Random Forest and GBM, will outperform traditional statistical methods in accurately predicting customer churn, with an accuracy rate exceeding 85% and an AUC-ROC of at least 0.90. Key predictors are anticipated to include contract type, monthly bill, data usage, customer tenure, and complaint history. The study is expected to demonstrate that integrating predictive analytics into customer management systems significantly enhances the ability to identify at-risk customers proactively, thereby enabling targeted retention strategies. This research contributes to the body of knowledge by empirically validating the application of machine learning models in churn prediction within the telecommunications sector, highlighting the interplay between customer behavior patterns and service variables. It extends existing literature by providing a comprehensive comparative analysis of multiple algorithms in a real-world operational context, offering actionable insights for industry practitioners. The main conclusion underscores the necessity for telecommunications firms to adopt data-driven, predictive frameworks to reduce churn and increase customer lifetime value. Based on the findings, recommendations include implementing real-time churn detection systems, customizing retention offers based on predictive insights, and continuously updating models to adapt to evolving customer behaviors. Future studies could explore the integration of text analytics from customer feedback and social media data, as well as assessing the impact of personalized retention interventions on reducing churn rates. This research paves the way for more sophisticated, analytics-driven customer management approaches essential for sustaining competitive advantage in the rapidly digitalizing telecommunications landscape.