Credit Card Fraud Detection using XGBoost Classifier with a Threshold Value

Abstract

Abstract Over the past years, speedy development of e-commerce techniques has been observed, making it promising for society to choose the best worthwhile product. This has made us dependent on financial institutions, where everyone deals with online banking facilities. Moreover, for payment, people are preferring credit cards over other methods which thus, have a higher risk of getting compromised. Thus, it is a big responsibility of financial institutions to upgrade their existing mechanism to prevent these fraud actions. However, it has also made it easy for scammers to exploit this big chance. Credit Card Fraud Detection helps us to identify fraudulent transactions. The proposed model in this paper detects fraud transaction using the XGBoost classifier to handle the imbalanced data. In the standard approach, the threshold value is pre-defined, which will lead to poor performance. Thus, in our proposed model, calculation and comparison of different threshold values are done to obtain the best value which gives an optimum result and high efficiency.