The Customer Churn Prediction with Artificial Intelligence Approaches in Banking Sector

Banks want to maintain ongoing relationships with their existing customers because it directly impacts their profitability. Long-term relationships with existing customers determine whether banks will continue to operate. In this context, banks use traditional and contemporary modelling approaches to predict and mitigate customer churn. The results obtained from these predictive studies are evaluated and used to inform actions taken by existing customers. The failure of existing customers to churn (maintaining loyalty) not only impacts the bank’s profits but also fosters the integration of new customers into the banking system by building robust customer base. In this research study, customer churn models were developed using artificial intelligence approaches for action. In this modelling process, customer-specific predictive results were obtained by completing preprocessing (transform, drop, type definition, etc.), hyperparameter tuning and supervised learning stages. The trained models were applied to both training and testing datasets to evaluate their performance. This study demonstrates that ensemble learning approaches (Random Forest and eXtreme Gradient Boosting) are highly successful in customer churn modelling. By deploying the developed models to bank processes, the churn status of existing customers can be predicted at certain intervals.