PREDICTION OF GLOBAL WARMING ON CROP YIELD INTEGRATING EDGE COMPUTING AND KALMAN RICCATI SUPPORT VECTOR REGRESSION

Abstract

Agricultural production to a large extent depends on weather conditions. The caution of global climate change or global warming has extensively caught the curiosity of researchers as these changes are reporting negative influences on overall crop production. Edge computing in the period of digital evolution, is steadily receiving drive across several industries. In spite of cloud configuration has hitherto played a paramount role in boosting the agricultural sector, edge computing outperforms in terms of both speed and efficiency. While exploiting edge computing technique farmers rely on data to acquire enhanced control over the industry and optimize the effectiveness of their functioning that in turn results in minimized operational expenses too. Edge computing and machine learning can be employed cooperatively in agriculture to enhance data processing and decision-making. In this work an edge computing model is integrated with machine learning technique for analyzing the impact of global warming on agriculture for enhancing crop yield prediction both accurately and in a computationally efficient manner. The method is called, Kalman Riccati Spatio-Temporal Support Vector Regression (KRST-SVR) based Crop yield prediction. With the overall method being split into three layers, the first layer collects the data and is sent to the second layer. The second layer or the edge device performs the pre-processing and feature extraction by employing Kalman Bucy Riccati Filter-based preprocessing and Spatial Temporal Regression based Feature Extraction. With this the edge device sends processed, relevant and valuable data to third layer. The third layer or the cloud server with the aid of Support Vector Regression function performs the classification with minimal error, therefore producing accurate and precise prediction results. Extensive simulation is done using Edge Cloud Simulator and Python programming language and detailed comparison is made with the state-of-the-art work in terms of prediction accuracy, precision, recall, prediction time and prediction error.