Machine learning approaches for credit card fraud detection: A comparative study of logistic regression and K-Nearest neighbors KNN

Abstract

The rapid expansion of online banking and digital transactions has revolutionized financial services, enhancing convenience and accessibility. However, this transformation has also led to an increase in fraudulent activities, posing significant risks to both consumers and financial institutions. Credit card fraud, in particular, has emerged as a prevalent threat, with fraudsters employing sophisticated techniques to exploit vulnerabilities in digital payment systems. Traditional fraud detection methods often fall short in identifying complex fraudulent patterns in real-time, necessitating the development of more robust and intelligent detection mechanisms. This study investigates the application of machine learning (ML) techniques, specifically Logistic Regression and K-Nearest Neighbors (KNN), to detect fraudulent credit card transactions. Using a dataset sourced from Kaggle, consisting of 568,630 transactions, the research evaluates the performance of these algorithms in accurately classifying transactions as fraudulent or legitimate. The results indicate that the KNN algorithm achieved higher accuracy on the evaluated dataset (0.999) compared to Logistic Regression (0.791). Similarly, KNN demonstrated superior classification performance across all evaluation metrics, including precision, recall, and F1-score. This study highlights the potential of machine learning in enhancing credit card fraud detection. The findings contribute to the ongoing efforts in financial cybersecurity, providing insights into the development of more effective and adaptive fraud prevention systems.