Stacked Ensemble-Based Machine Learning Approach for Anomaly Detection in Credit Card Transactions

The growing prevalence of digital transactions has unfortunately brought along an increase in fraudulent activities, particularly in the credit card transactions. To counter this, our research explores the use of a stacked ensemble model in the detection of credit card anomaly. Our proposed model leverages algorithms, such as random forest, gradient-boosted tree and logistic regression which allows the model to exploit on the strength of individual base model, improving the overall predictive performance. To test the effectiveness of the stacked ensemble model, we used BankSim credit card transaction datasets, one of which was highly imbalanced, a common characteristic of fraud detection datasets where fraudulent transactions is vastly outnumbered by legitimate ones. This objective is where our stacked ensemble model could leverage the diversity of the base models to navigate the class imbalance and to improve the accuracy and reliability of credit card anomaly prediction. The performance of proposed model is compared with majority voting ensemble (MVE) (Al-Faqir in J Theor Appl Inf Technol 100(14), 2022 [1]) model.