JoWUA

Abstract

The high-speed, high-volume data used in edge computing is a fast-changing world, so the safe storage and conversion of the data to generate real-time decisions and analytics is essential. In this paper, a new solution to transforming, storing, and analyzing real-time data streams is suggested in edge computing environments. The process starts by obtaining stock data that is financial in nature followed by cleaning and normalization of the data by the Adaptive Two-Stage Unscented Kalman Filter (ATUKF). The processed data is then clustered in order to detect underlying patterns and cluster like points. The Finite Basis Physics-Informed Neural Networks (FBPINN) approach is used to optimize service interactions and minimise service response time and delays during data transmission in order to reduce latency. The method is executed in Python and experiments prove it to have a 99.2 % accuracy with a computation time of 1.150 seconds. The suggested approach is much better than the current approaches, such as Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and Deep Neural Networks (DNN). This method can be used in high-frequency trading and other applications that need real-time finances data processing, as it is a more efficient and scalable solution that can be achieved by utilizing edge computing and sophisticated machine learning methods. The findings demonstrate the capability of combining ATUKF to process the data, clustering to identify the patterns, and FBPINN to optimize the response of the service as an overall structure to improve the performance of the edge-based real-time data analytics.