An Innovative Method for Fraud Detection in E-Commerce Using DCNN-Multiclass SVM Model

A growing number of monetary transactions are taking place through different E-commerce platforms in this age of big data. Theft of personal information for fraudulent reasons is one of the risks that could arise from these opportunities. Credit cards are a prime target for cybercriminals seeking to steal personal data or execute fraudulent transactions due to their extensive use for online shopping. Intelligent fraud detection systems have come a long way, yet the system still haven't solved the notorious problems brought on by data imbalances. Following this important sequence will ensure that the data is properly prepared for feature extraction and model training. Prior to ATF processing raw log files pertaining to online e-commerce activity, a preprocessor removes data noise, such as click records created by customers who have not registered. Data mining for features an evolving metaheuristic algorithm called the multi-verse optimizer imitates the laws of the popular theory called multi-verse. While training the model, the proposed approach utilized DCNN-Multiclass SVM. Two state-of-the-art approaches, CNN and SVM, are beaten by the suggested approach. After using the strategy, accuracy improved by 95.65%.