Innovative Fusion of Local Outlier Factor and Isolation Trees for Advanced Credit Card Fraud Detection

Abstract

This study proposes a combination of the local outlier factor algorithm and isolation trees to detect credit card fraud efficiently. The approach involves dataset pre-processing, outlier identification using isolation trees, and individual scoring with a local outlier factor. Despite the challenge of distinguishing fraudulent transactions, the method demonstrates a low false positive rate, high accuracy, speed, and adaptability, making it suitable for advanced fraud detection systems. Its applications extend to the financial sector, mitigating losses from credit card theft, and potentially aiding fraud detection in other domains. Future research will focus on enhancing the approach with additional machine learning algorithms and evaluating their effectiveness on diverse credit card fraud datasets. In summary, the proposed technique holds promise for reducing false positives while maintaining accuracy, contributing to the field of fraud detection.