Handling Data Imbalance Using a Heterogeneous Bagging-Based Stacked Ensemble (HBSE) for Credit Card Fraud Detection

Increase in electronic payments has lured fraudsters into this domain, leading to reduced security for customers. Although banks continuously strive to provide enhanced security for the prevention of frauds, these are at times countered by the fraudsters. This has become a vicious cycle leading to the requirement of fraud detection model rather than a fraud prevention mechanism. This work presents a heterogeneous bagging-based stacked ensemble (HBSE) model for effective detection of fraud in credit card transactions. The model is composed of a bagging mechanism that aims to handle data imbalance and heterogeneous base learners that act as effective decision rule extraction mechanisms, hence effectively operating on the highly complex data. The process of creating an ensemble ensures that the model is interpretable. Experiments were conducted on benchmark data, and comparisons were performed with state-of-the-art models. Comparisons indicate highly effective performance of the proposed HBSE model.