Abstract—The number of online banking and financial services through mobile apps is growing steadily. The number of customers using these apps for their monetary transactions is drastically increasing day by day. However, the increasing use of these apps on smart devices raises security concerns. Therefore, it is becoming a significant process to implement effective mechanisms to prevent fraud and protect personal data. In today’s financial world, the use of credit cards for online purchases has increased exponentially, and with it, the fraud that accompanies it. It is very difficult to detect fraudulent transactions in banking transactions. Therefore, the proposed work has introduced a novel deep learning framework based on a complexity-reduced Multi-Octave Dilated Temporal Convolutional Network (MO-DTCN) with Explainable Artificial Intelligence (XAI) approaches, including Lime and Shap. With these advancements, the proposed model can detect fraudulent financial transactions. To reduce the complexity of the architecture, the proposed work has utilized a Weighted Pruning and Quantization approach. In addition, meta-learning helps to fine-tune the performance model. Experimental evaluation on this dataset proves that the proposed MO-DTCN framework is effective, as it attains an accuracy of 99.13%, precision of 99.48%, recall of 99.35%, and F1-Score of 99.60% for an 80% training split with a very high Matthews Correlation Coefficient (MCC) of 0.9945, reflecting strong robustness on highly imbalanced data.
 
Keywords—Multi-Octave Dilated Temporal Convolutional Network (MO-DTCN), financial fraud detection, Explainable Artificial Intelligence (XAI), meta-learning, lightweight deep learning models


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).