A Hybrid Deep Learning And Modified Butterfly Optimization Based Feature Selection For Transaction Credit Card Fraud Detection

Abstract

Credit cards are playing an extremely significant part in the modern economy. However, as the number of people who use credit cards continues to climb, the number of fraudulent credit cards transactions has also increased. There are numerous different ideas put up in order to combat the rising incidence of credit card theft. In the existing research work, developed an ENNs (Enhanced Neural Networks) for enhanced accuracy of results using feature selection techniques based on ABCs (Artificial Bee Colonies) which select relevant features from transaction level credit card datasets. Therefore, the quality of the classification will vary based on the input data dimension only and the expense of making accurate decisions grows to be a serious issue. So this research work, introduced a hybrid deep learning and adaptive feature selection methodology for efficiently detecting credit card frauds. First, the characteristics of the transactional documents need to be completely ordered, and then the contents of each feature need to be categorized. Construct a Logical Graph of Behavior Profile (LGBP) using them as a foundation. This graph should abstract and include all distinct transaction data. Then the Modified Butterfly Optimization Algorithm (MBOA) based feature selection has been utilized in order important to identify characteristics from transaction level credit card dataset. And finally the hybrid deep learning model is used in order to depict the rational connection between the characteristics of transactional details. Here the Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) is hybridized for improving the detection performance of the transactions. According to the findings of the simulations, the suggested hybrid deep learning model has a higher reliability and recognition rate than the other models that are currently available.