Detecting Credit Card Fraud: A Comparative Analysis of KNN, Random Forest, and Logistic Regression Methods

In the dynamic field of digital finance, credit card fraud poses a significant challenge, undermining transaction security and consumer confidence. This research paper evaluates the efficacy of three prominent machine learning algorithms—K-Nearest Neighbors (KNN), Random Forest (RF), and Logistic Regression (LR)—in detecting credit card fraud, utilizing a rigorous comparative analysis based on accuracy, precision, recall, and the F1-score. Employing a comprehensive dataset representative of real-world transactions, the study aims to expose the unique strengths and operational efficiencies of these algorithms. The results particularly highlight the performance of the KNN algorithm, which was meticulously trained and validated on a normalized and partitioned dataset. The optimal ’k’ value was identified through cross-validation to prevent overfitting and underfitting, enhancing its real-world applicability. With a Euclidean distance metric, the KNN model demonstrated high effectiveness, achieving an optimal balance between precision and recall with a ’k’ value of 5, leading to an F1-score of 0.91 and accuracy of 0.96. These findings not only affirm the capabilities and limitations of KNN, RF, and LR in fraud detection but also set the stage for future enhancements in machine learning strategies to combat credit card fraud efficiently.