Vasant, Mohit, Ganesan, Swathi ORCID: https://orcid.org/0000-0002-6278-2090 and Kumar, Ganapath (2025) Enhancing E-commerce Security: A Hybrid Machine Learning Approach to Fraud Detection. FinTech and Sustainable Innovation, 1.

Preview

Text 08-FSI52024882_online+revised.pdf - Published Version Available under License Creative Commons Attribution. Download (946kB) | Preview

Abstract

In the rapidly expanding e-commerce landscape, ensuring the security of transactions is essential to maintain consumer trust. However, the challenge of accurately distinguishing between genuine and fraudulent transactions persists, largely due to issues such as dataset imbalance, suboptimal feature selection, and varying algorithm performance. As such, this study aims to enhance fraud detection accuracy by developing a hybrid model that combines an artificial neural network (ANN) with a deep neural network (DNN), employing the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance and linear discriminant analysis (LDA) for effective feature extraction. By integrating SMOTE with LDA, the model is trained to better handle imbalanced datasets and extract relevant features, thereby improving its predictive capabilities. Our results demonstrate that the hybrid model outperforms individual models, achieving a precision rate of 95.46% and an area under the curve (AUC) score of 97.04%. In comparison, the stand-alone ANN model recorded an accuracy of 95.46% and an AUC of 96.92%, while the DNN achieved a success rate of 95.01% and an AUC of 97.17%. These outcomes highlight the significant advantages of combining advanced feature extraction and class imbalance techniques, resulting in superior detection performance. The study concludes that the hybrid model provides a robust solution for improving fraud detection in e-commerce, offering a reliable approach to differentiate between genuine and fraudulent transactions effectively. This approach not only addresses existing challenges but also sets a foundation for future research in enhancing transaction security through innovative deep learning methodologies.

Item Type:	Article
Status:	Published
DOI:	10.47852/bonviewFSI52024882
School/Department:	London Campus
URI:	https://ray.yorksj.ac.uk/id/eprint/12669

University Staff: Request a correction | RaY Editors: Update this record