Comparative Evaluation of Machine Learning Algorithms for Detecting Credit Card Fraud

Detecting credit card fraud is an extremely important issue in the financial sector, as it results in significant financial losses and negatively impacts customer trust. The main motive of CCFD is to develop methods to identify fraudulent transactions accurately and efficiently. This research typically involves the use of statistical and machine-learning algorithms to analyze large amounts of transaction data. Utilizing machine learning is a potential solution for addressing credit card fraud. Studies suggest that machine learning techniques can help overcome the challenges of identifying fraudulent transactions with high detection rates, both directly and indirectly. While supervised ML algorithm and unsupervised ML algorithms have been proposed, the limitations of each approach underscore the need for hybrid methods. From the results, it is evident that Logistic Regression has the highest accuracy of 94.86%, highest recall value of 98.97% and highest F1-score of 95.09%.

I. Introduction

As the globe transformed into a cashless society, dependency on internet transactions will increase. E-payment systems are critical in today's financially competitive culture. The use of credit cards has simplified the process of buying products and services. Financial institutions provide their customers with cards that enable them to make purchases without the need for cash. Credit cards offer several advantages over debit cards, such as safeguarding customers against damaged, lost, or stolen goods. Before initiating any transaction with their credit card, customers are required to verify the purchase with the retailer.