Enhancing Financial Fraud Detection with LSTM and Attention Mechanisms: A Real-Time, Big Data Approach

Electronic checks and consequent fraud cases increase as internet activity increases. In real-time, it uses deep learning based on LSTM networks and the attention processes to detect fraud. Big and continuous datasets assist the model in identifying unusual behavior. Thus, the performance of the proposed approach was examined using 1,000,000 financial transactions. For precision, it yielded 96.7 percent, recall achievement was 97.4 percent, accuracy setting was 98.2 percent, and F1 score was 97.0 percent. Overall accuracy classification of 98.5% and false positive of 1.8%. The model was computationally efficient since it only took five hours to train and only 2 milliseconds to test every transaction. From the following findings, it is clear that the illustrated system deletes possibilities of financial fraud, and offers efficient, real-time, and extendable banking solutions. Future work will enhance the generalizability and the interpretability of the model.