Customer Engagement Prediction in Retail Banking with Explainable AI

Purpose: In the highly competitive retail banking environment, accurately predicting customer engagement is critical to enhancing customer satisfaction, ensuring retention, and ultimately boosting profitability. Materials and Methods: This study embarks on a comprehensive exploration of advanced computational finance techniques by integrating Explainable Ensemble Learning (EEL) with a suite of Trustworthy Open AI tools. Utilizing state-of-the-art methods including Evidently AI for rigorous testing and LIME for interpretability this research leverages the publicly available Berka dataset. Findings: After an intensive phase of feature engineering, extraction, and deep clustering, customers are segmented into distinct engagement categories. Four ensemble models are constructed and meticulously evaluated, with the blending model emerging as the most effective approach by achieving an impeccable AUC score of 1.000 along with outstanding accuracy, precision, recall, and F1 scores. Implications to Theory, Practice and Policy: Only six out of 2000 data points were misclassified, underscoring the model’s robustness. This paper not only highlights significant advancements in using computational finance techniques to predict customer engagement but also emphasizes the crucial role of transparency and interpretability in fostering trust within AI-based decision systems.