The volume of payment transactions has grown exponentially, creating a high demand for high-throughput payment processing systems. These systems must be capable of handling a large number of transactions with minimal delay while also being highly scalable and resilient to failures. To overcome this challenge, leveraging kafka for event-driven architecture in fintech applications (LK-EDA-FA-BSCNN) is proposed. At first, input data is gathered from kafka streams. Then, the input data are pre-processed using adaptive two-stage unscented kalman filter (ATSUKF is used to clean the data to ensure high-quality input for downstream analysis. Then, the pre-processed data are fed into binarized simplicial convolutional neural network (BSCNN) is used to predict the future transactions from historical trends. The proposed LK-EDA-FA-BSCNN method is implemented using python and the performance metrics like accuracy, precision, sensitivity, specificity, F1-score, and computational time. The LK-EDA-FA-BSCNN method achieves the best performance with 98.5% accuracy, 95.3% precision and 1.150 seconds runtime with existing methods, like a DRL-based adaptive consortium blockchain sharding framework for supply chain finance (DRL-ACSF-SCF), a blockchain-based secure storage and access control scheme for supply chain finance (BC-SS-ACS-SCF), and analysis of banking fraud detection methods through machine learning strategies in the era of digital transactions respectively.
Modadugu et al. (Sat,) studied this question.