Leveraging Machine Learning Algorithms for Robust Credit Card Fraud Detection: A Performance Comparison

The implementation of systems designed to detect cyber crimes has become essential for banks and other financial institutions. A variety of approaches have been proposed to combat the increasing incidence of credit card fraud. Fraud detection entails the analysis of user or customer purchasing behavior to assess, identify, or prevent undesirable activities. This paper aims to employ several machine learning (ML) algorithms, including Logistic Regression, Decision Trees, Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbor, to ascertain the legitimacy of transactions based on their value. Initially, a more balanced dataset is created through the random undersampling technique, followed by the application of grid search to optimize the parameters utilized in the various ML algorithms. A comparative analysis of the performance of these ML algorithms is conducted, revealing that the Random Forest method is particularly effective in identifying anomalous patterns within transaction data.