Probabilistic Credit Card Fraud Detection System in Online Transactions

Abstract

This paper discussed the past works on fraud detection system and highlights their deficiencies. A probabilistic based model was proposed to serve as a basis for mathematical derivation for adaptive threshold algorithm for detecting anomaly transactions. The model was optimized with Baum-Welsh and hybrid posterior-Viterbi algorithms. A credit card transactional data was simulated, trained and predicted for fraud. And finally, the proposed model was evaluated with different metric. The results showed that with the optimization of parameters, posterior-Viterbi cum new detection model performed better than Viterbi cum old detection model.