Bidirectional gated recurrent unit for improving classification in credit card fraud detection

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