ChurnNet-XAI: A Neural Network-Based Framework for Customer Attrition Prediction in Banking

The attrition of customers is a major problem in the financial services industry that has a significant impact on the institutional profitability and long run sustainability. Timely recognition of those customers that have high chances of abandoning the services allows the organizations to adopt proactive retention strategies and rational resource allocation. The classical statistical and machine learning models like the Logistic Regression and the Support Vector Machines have been extensively utilized in churn prediction, but cannot easily characterize non-linear relationships between variables in large customer data. The recent developments in the field of deep learning have shown better results on the high-dimensional behavioral data modeling owing to their capability to harness the complex interaction of features. This paper will suggest an intelligent predictive model which is Deep Neural Networks (DNN) with Explainable Artificial Intelligence (XAI) to predict churn in banking settings. The model is conditioned on structured customer information such as demographic and transactional behavior, account tenure and financial metrics. To improve predictive reliability, preprocessing, like feature normalization and control of the imbalance of classes with SMOTE, is included.