Credit Card Fraud Detection with NCA Dimensionality Reduction

Abstract

Credit card transactions for online payments have increased dramatically and fraud attempts on these payments have become prevalent with more advanced attacks. Thus, conventional fraud detection mechanisms are inadequate to provide acceptable accuracy for fraud detections. Machine learning algorithms may provide a proactive mechanism to prevent credit card fraud with acceptable accuracy. In paper, we propose a new approach with machine learning for credit card fraud detections by increasing the performance of classification algorithms. We use the Neighborhood Component Analysis (NCA) dimensionality reduction to improve success rate for credit card fraud detections that use K-Nearest Neighbors (KNN) classification algorithm. We implemented the proposed approach and we tested it on a dataset. Particularly, we evaluated the results with the Area under the Receiver Operating Characteristic (AUROC) metric. The analyses results show that our approach provides better accuracy for credit card fraud detections.