Comparative Analysis of Fraud Detection of Credit Card using Supervised and Unsupervised Learning

The unlawful use of a credit card or other such payment device to obtain cash or property without authorization is known as credit card fraud. Fraudsters steal credit card details from unprotected websites or acquire them through identity theft schemes. A Fraud Detection System (FDS) is implemented to identify fraudulent transactions. With the increasing risk of credit card fraud, where criminals exploit credit cards to access funds from others’ accounts, detecting fraud in credit card transactions has become a critical area of research. In recent years, machine learning has demonstrated positive results when it comes to identifying credit card fraud. In order to evade detection by the present fraud detection systems, fraudsters refine their methods, and a plethora of proposed machine learning models exist that may detect fraud cases using a certain dataset. Though each model functions well to some degree, not all of them are the greatest. This study offers a comprehensive overview of various models by presenting a comparative analysis of credit card fraud detection utilizing both supervised and unsupervised learning approaches, as numerous machine learning models are currently available for this task. This represents the first comparative evaluation of eighteen different models. The performance of all models was assessed using metrics such as precision, recall, and F 1 -score, alongside the plotting and computation of ROC curves and AUC scores for comparison. This paper can be used as a research tool because it compares about eighteen different models.