Gradient Boosting Techniques for Credit Card Fraud Detection

Abstract

Now-a -Days e-commerce has enabled increased online transactions, and hence growing serious credit card frauds. This malicious activities are affecting millions of people' identity theft and loss of money.  The fraudsters are continuously adopting new ways to perform illegal activities. This paper gives a detailed analogy of different supervised and unsupervised machine learning techniques for detecting fraudulent activities. The new schemes Cat Boost and Light Gradient Boosting Machine (LGBM) are proposed for fraud discovery. The performance of these methods is compared with approaches of Auto Encoder (AE), Logistic Regression and K-Means clustering and Neural Network (NN)  and found that Cat Boost and LGBM are giving high accuracy in fraud detection.