An Efficient Machine Learning and Filter Based Implementation of Data Fusion in WSN

Abstract

Due to the large number of datasets produced by various sensors integrated within various IoT (Internet ofThings) applications worldwide, multiple-sensor dataset fusion has become a major challenge in the modernera. In recent years, researchers have developed a variety of solutions for improved data fusion processes for reliable data processing in the IoT and wireless sensor networks (WSNs) environments. However, such existing models have a variety of limitations due to limited design constraints for data processing in IoT-based WSNs. This has indeed been extensively presumed as just the robust non-linear system because of high computational complexities generated in response within the entire functioning. A meticulous and appropriate methodological solution is becoming a difficult task to accomplish. In order to address theaforementioned issues, the authors of this article created an improved model by combining the ML(Machine-Learning) algorithm with the Kalman filter for a more accurate and precise centralized datafusion process in the WSNs environment. Furthermore, our developed model is more energy efficient thanprevious models due to its lower computational complexity design. In comparison to previous models, theresults of the proposed model indicate a gradual improvement in overall prediction accuracy. The proposed model includes precision values of 97.98 percent, 95.12 percent, 97.18 percent, and 97.84 percent for accuracy, F1-scoring, and recall, among other parameters. In terms of performance, all of these parametersoutperform the previous model. More investigation into this area for performance simulation of the system on high volume data sources of sensors in the WSNs scenario is possible in the future.