Credit Card Fraud Detection in Data Mining u sing XGBoost Classifier

Abstract

In today's economy, credit card (CC) plays a major role. It is an inevitable part of a household, business & global business. While using CCs can offer huge advantages if used cautiously and safely, significant credit & financial damage can be incurred by fraudulent activity. Several methods to deal with the rising credit card fraud (CCF) have been suggested. Both such strategies, though, are meant to prevent CCFs; each of them has its own drawbacks, benefits, and functions. CCF has become a significant global concern because of the huge growth of e-commerce and the proliferation of payment online. Machine learning (ML) algo as a data mining technology (DM) was recently very involved in the detection of CCF. There are however several challenges, including the absence of publicly available data sets, high unbalanced size, and different confusing behavior. In this paper, we discuss the state of the art in credit card fraud detection (CCFD), dataset and assessment standards after analyzing issues with the CCFD. Dataset is publicly available in the CCFD data set used in experiments. Here, we compare two ML algos of performance: Logistic Regression (LR) and XGBoost in detecting CCF Transactions Real Life Data. XGBoosthas an inherent ability to handle missing values. When XGBoost encounters node at lost value, it tries to split left & right hands & learn all ways to the highest loss. This is when the test runs on the data. The experimental results show an effective use of the XGBoost classifier. Technique of performance is widely accepted metric based on exclusion: accuracy & recall. Also, the comparison between both approaches displayed based on the ROC curve.