Improving E-Commerce Fraud Detection: A GAN and Reinforcement Learning Approach Integrated with Personality Analysis for Secure Digital Economy

The rapid growth of online e-commerce makes platforms more vulnerable to fraud. Fraudulent behavior hurts e-commerce rating systems and user experiences. Fraudsters change their approaches to avoid detection using complicated adversarial methods. This requires efficient and effective mechanisms to identify and control such actions. Preprocessing, feature extraction, and model training comprise the suggested method. During the preprocessing step, duplicates are removed, repeated matches are ignored, and finally, the remaining pairs of persons are joined. Combining word2vec and doc2vec with TF-IDF produces helpful feature extraction representations. A RL-GAN is used to trains the model. The suggested RL-GAN model outperformed solo RL and GAN models. Its 93.49% accuracy rate shows its effectiveness in detecting e-commerce fraud. Results reveal that the suggested strategy can resist new online fraud. The RL-GAN model uses cutting-edge preprocessing, feature extraction, and training to detect e-commerce fraud.