Bidirectional Gated Recurrent Unit for improving classification in credit card fraud detection

Abstract

The volume of credit card transactions is increasing considerably in recent years. Thus, the number of fraud cases have also increased, resulting in losses of thousands of dollars worldwide. Therefore, it is mandatory to use techniques that are able to assist in the detection of credit card fraud. For this purpose, we have proposed a multi-layer architecture, composed of four layer: authentication layer, behavioural layer, smart layer and background processing layer. In this paper, we focus on the implementation of the smart layer. The aim of this layer is to develop a model for the detection of credit card fraud, using bidirectionel gated recurrent units (BGRU).  The experiments, applied on well-known credit card fraud dataset from Kaggle, show that this model has peak performance compared to other proposed models, with 97.16% for accuracy rate and 99.66% for the area under the ROC curve.