TCCFD: An Efficient Tree-based Framework for Credit Card Fraud Detection

Due to digitalization, credit card fraud has become one of the most prevalent global risks. The global economy loses billions of dollars annually due to credit card fraud. Financial institutions adopt a strategic approach to credit card fraud mitigation, relying on business rules to combat the occurrence of fraud and sophisticated analysis to detect fraud. However, Fraudsters are altering their patterns regularly with the help of new technologies. Thus, organizations continually seek to detect new fraud patterns. The detection of credit card fraud has improved tremendously through using artificial intelligence, particularly machine learning techniques. We introduce through this paper an efficient framework that adopts various tree-based machine learning models to detect fraudulent transactions. We apply the proposed framework to a highly asymmetric real-world dataset. The proposed framework integrating the extra trees classifier with the borderline synthetic minority oversampling technique surpasses others and previously developed models with 99.96 % accuracy and 96% area under the ROC curve. The proposed framework asserts its efficacy and reliability in classification while also addressing the class imbalance issue.