Detection of Fraudulent Credit Card Transactions Using Advanced LightGBM Approach

Abstract

Due to the rapid growth of e-commerce businesses, the use of credit cards for online purchases has increased exponentially and so has fraud. In recent years, it has gotten increasingly difficult for banks to detect fraud in credit card systems. Hence, it is important to have effective and reliable methods for identifying fraud in credit card transactions. Machine learning is critical for the detection of fraudulent credit card transactions. Banks utilize various machine learning approaches to forecast these transactions; prior data are collected, and new features are applied to improve prediction capability. Classical algorithms such as Logistic Regression (LR) and Random Forest (RF) are proven useful for detection of fraudulent credit card transactions. However, their performance is not satisfactory. Hence, we proposed an approach to improve the detection accuracy. This paper introduces an Advanced Light Gradient Boosting Machine technique to detect credit card frauds. In this method, a Bayesian-based Hyperparameter Optimization technique is incorporated to Light Gradient Boosting Machine (LightGBM) in modifying its hyperparameters. To validate its efficacy in identifying fraudulent credit card transactions, the proposed Advanced LightGBM technique was evaluated on two credit card transaction datasets that comprised both fraudulent and genuine transactions. The method outperformed current techniques in terms of performance.