Paper Contents
Abstract
The exponential growth of digital financial transactions has significantly increased the risk of online payment fraud, posing major challenges to the integrity and security of e-commerce systems. This research addresses the problem by exploring and evaluating deep learning models for the detection of fraudulent transactions. The study utilizes a large, real-world Kaggle dataset of online payments and applies advanced data pre-processing techniques to manage class imbalance and ensure data quality.Three deep learning models Feedforward Neural Networks (FNN), Long Short-Term Memory (LSTM) networks, and Auto Encoders are implemented and evaluated using performance metrics such as accuracy, precision, recall and Mean Absolute Error (MAE). The FNN achieved the highest accuracy of 99.41%, followed closely by LSTM with 99.37%, while the Auto Encoder, though less accurate at 85.20%, demonstrated potential for unsupervised anomaly detection. The results reveal that supervised learning models (FNN and LSTM) outperform unsupervised models in binary fraud classification, with FNN proving most effective for real-time transaction analysis.
Copyright
Copyright © 2025 Dr. Neeru Ahuja. This is an open access article distributed under the Creative Commons Attribution License.
