Anomaly Recognition in Digital Commerce for Unethical Activity Prevention Using Advanced Learning Models

MCA

This research introduces a system titled "Adaptive Illicit Transaction Identification in Online Marketplaces via Computational Intelligence", which focuses on the detection of fraudulent activities within e-commerce platforms. The system architecture is developed using Python for backend processing and employs HTML, CSS, and JavaScript on the frontend, integrated seamlessly through the Flask web framework to ensure a dynamic and user-friendly interface. To achieve high detection accuracy, the system incorporates two sophisticated machine learning models: a Stacking Classifier and an XGBoost (XGB) Classifier. The Stacking Classifier recorded a perfect training accuracy of 100% and a test accuracy of 99%, while the XGB Classifier attained 96% training and 95% test accuracy. These outcomes highlight the models’ robustness in effectively distinguishing between legitimate and fraudulent transactions. A synthetic dataset of 23,634 records was generated using the Faker library, enhanced with custom logic to realistically simulate transaction behavior and illicit patterns. The dataset features 16 key attributes—including transaction amount, payment method, and fraud indicators that collectively support the accurate modeling of transactional anomalies. The results confirm the potential of computational intelligence and ensemble learning approaches in strengthening fraud detection mechanisms, thereby enhancing transactional security and user trust in online marketplaces

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.