Comparative Evaluation of Fraud Detection in Online Payments Using CNN-BiGRU-A Approach

Online payment systems are now ubiquitous in people's daily lives. However, with this increased convenience comes a new security risk. Due to the nature of cybercrime involving an online payment service, which is often characterized by specialization, diversification, concealment, industrialization, scenario, and cross-regional elements, online payment security is notoriously difficult to implement and maintain. Effective and comprehensive online payment fraud detection is crucial. The first three stages of the proposed technique are preprocessing, feature selection, and model training. Remove Null Value, Data Transform, Resample, and Normalize Data are all preprocessing techniques used to improve the quality of raw data. Feature selection makes use of PCA and IG as a means of analysis. CNN-BiGRU-A is used to train the models after information gain is used to choose relevant features. When compared to the two most common alternatives, BiGRU and CNN, the proposed technique performs better.