Optimized churn prediction using ensemble-based feature selection via second-order cone programming

This paper investigates the diverse applications of ensemble-based artificial intelligence algorithms in analyzing customer churn within the banking industry. Additionally, it conducts a comprehensive comparison, rigorously contrasting well-established conventional methods with the relatively less-familiar second-order cone programming (SOCP) ensemble-based feature selection method. Conducting a comprehensive analysis, the study explores large-scale feature engineering using demographic, financial and transactional data. This approach not only enhances the complexity and real-world adaptability of churn analysis methodologies but also makes a significant contribution to business analytics in the banking sector. The research stands as a rare example in the literature, utilizing bank customer data within the interest-free financial system, thereby advancing our understanding of customer relationship management in both the operations research and financial sectors. Banks can use these methods to enhance their customer retention strategies, which may result in lower churn rates and higher customer lifetime value. By minimizing customer churn, banks can improve their financial stability and profitability, potentially leading to more consistent lending practices and lower interest rates for customers.