Abstract

Customer churn is still a big problem for the retail industry because it affects profits and long-term survival. Machine learning (ML) has become a powerful tool in the last few years for predicting and preventing churn by finding customers who are likely to leave and allowing data-driven retention strategies. This paper examines the utilisation of machine learning techniques for predicting customer churn in the retail sector, integrating results from previous research and suggesting a comprehensive methodological framework for effective implementation. The study analyses 30 peer-reviewed journal articles published until 2023, focusing on significant machine learning models like logistic regression, decision trees, random forests, gradient boosting machines, and deep learning techniques, emphasising their relative advantages and disadvantages in churn prediction scenarios. The methodology section outlines a systematic pipeline that encompasses data preprocessing, feature engineering, model training, evaluation metrics, and deployment considerations. Results from empirical implementations across diverse retail datasets demonstrate that ensemble methods (e.g., XGBoost, LightGBM) and neural networks consistently outperform traditional statistical models in predictive accuracy, though model interpretability remains a critical concern for practitioners. The paper also talks about how explainable AI can help people trust and use ML-driven churn prediction systems. The results highlight the necessity of incorporating ML models with customer relationship management (CRM) systems to facilitate prompt interventions, tailored offers, and the enhancement of loyalty programs. This study enhances academic scholarship and industry practice by addressing methodological, practical, and ethical considerations. The paper ends with suggestions for retailers who want to use ML tools and ideas for future research.