Lightweight Deep Learning Approach Using 1D-CNN and Attention for Sequential Credit Card Fraud Detection

Fraudulent activity in credit card transactions continues to be a pressing concern in the financial industry, primarily because transaction data is highly complex and heavily skewed toward legitimate cases. To address this issue, the present study proposes a hybrid deep learning framework that merges the strengths of a one-dimensional convolutional neural network (1D-CNN) with the selective capabilities of an attention mechanism. The performance of this enhanced model was rigorously compared with a conventional 1D-CNN, employing widely recognized evaluation metrics such as accuracy, precision, recall, and the F1-score. The experimental outcomes demonstrate that introducing the attention layer substantially improves the network’s ability to recognize critical temporal dependencies in transaction sequences. As a result, the model achieved exceptional performance levels, with an accuracy of 98%, precision of 97%, recall of 98%, and an F1-score of 98%. These findings provide strong evidence of the superiority of the attention-based approach, highlighting its effectiveness in producing more reliable and resilient fraud detection systems. Beyond the algorithmic gains, the research contributes a practical foundation for real-time applications in financial security, enabling institutions to curtail potential losses, reinforce public confidence in digital payment services, and enhance the efficiency of day-to-day operations.