Implementation of Stacking Ensemble Learning for Bank Term Deposit Acceptance Classification

Accurately classifying bank term deposit acceptance is critical for optimizing marketing strategies. This study proposes a novel Stacked Ensemble Learning (SEL) approach to improve classification accuracy in this domain. The proposed SEL leverages six diverse base learners namely Decision Tree (DT), Random Forest (RF), Gradient Boosting (GB), XGBoost (XGB), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) to exploit their complementary strengths and mitigate individual weaknesses. Each base learner generates predictions that are then used as features to sequentially train the six algorithms as meta-classifiers. This iterative process aims to harness the collective knowledge of all base learners. Evaluated on a resampled Bank Marketing Dataset using ENN undersampling, the SEL approach achieved its highest performance utilizing Random Forest as the final meta-classifier, with an accuracy of 96.75%, an F1-score of 96.72%, and an AUC-ROC of 0.9038. These findings highlight the effectiveness of SEL for bank term deposit classification. Our work contributes to the field by exploring a novel SEL architecture and demonstrating its success in a financial context. Future research can investigate alternative ensemble architectures and explore the factors influencing the performance of specific algorithms within this domain.