A two-phase feature selection technique using mutual information and XGB-RFE for credit card fraud detection

In 2025, Nilson report pointed out that the gross credit card fraud worldwide has been expected to be $35.31 billion [1]. It was found successful in fighting against criminals by Machine Learning (ML) models through analysing massive datasets generated, but still, the happening of fraud cannot be stopped The high volume of transactions needed to be processed to identify the fraud that does not happen frequently generating an imbalanced dataset The fraud considered as a legitimate transaction is the considerably higher cost than identifying a legitimate transaction as fraud As e-commerce widely grows, merchants are charged back for fraud loss generated