A Proposed Model for Card Fraud Detection Based on CatBoost and Deep Neural Network

The rapid development of technology has digitized customer payment behavior towards a cashless society. To certain extent, this created feast for miscreants to commit fraud. According Nilson (2020), global fraud loss is projected reach over $\$ $ 35 billion by 2025. Consequently, the need novel method prevent menace undisputed. This research was conducted on IEEE-CIS Fraud Detection Dataset provided Vesta Corporation. Based logic labeling converting entire account “Fraud=1” once credit card fraud, we navigate process predicting fraudulent cards rather than transactions. key idea behind proposed model user separation, in which divide users into old and new people before applying CatBoost Deep Neural Network each category, respectively. In addition, variety techniques improve detection accuracy, namely handling heavily imbalanced datasets, feature transformation, engineering, are also presented detail paper. experimental results showed that our performed well, as obtained AUC scores 0.97 (CatBoost) 0.84 (Deep Network).