Managing Credit Card Fraud Risk by Autoencoders : (ICPAI2020)

This paper introduces a risk control framework on credit card fraud instead of providing a solely binary classifier model. The anomaly detection approach is adopted to identify fraud events as the outliers of the reconstruction error of a trained autoencoder. The trained autoencoder shows the well fitness and robustness on the normal transactions and heterogeneous behavior on fraud activities. The cost of false positive normal transactions is controlled and the loss of false negative frauds can be evaluated by the thresholds from the percentiles of reconstruction error of trained autoencoder on normal transactions. To align the risk assessment of economical and financial estimation, Risk manager can adjust the threshold to meet the risk control requirements. Using 95th percentile as the threshold, the rate of wrongly detecting normal transaction is controlled at 5% and true positive rate is 86%. For 99th percentile threshold, the well controlled false positive rate is around 1% and 83% for the truely detecting fraud activities. The performance of false positive rate and true positive rate are competitive with other supervised-learning algorithms.