Two Level Adaptive Decision Support System For Crop Yield Prediction In Edge Cloud Enabled Smart Farming Environment Using Hybrid Deep Learning Approach

Abstract

In smart agriculture, the crop yield optimization and management is highly challenging due to its adaptive environmental conditions and wide data generation. Conventional crop yield management techniques are lacked with real time data processing and precise crop yield prediction on dynamic climatic conditions with timely recommendations. To this end, we design a novel Two Level Adaptive Decision Support System (TLA-DSS) for robust crop yield management using IoT sensors and UAV in edge-cloud enabled smart farming environment. The entities involved in the proposed work such as IoT sensors, UAVs, Distributed Edge Server (DES), 6G base station, and Centralized Cloud Server (CSS). At first, the smart agricultural data from the IoT sensors and UAV are pre-processed in terms of noise reduction and Vegetation Index (VI) computation in the DES. Along with the pre-processed data and climatic data information from the weather stations are utilized for Climatic Anomaly Detection (CAD) from Multi Head Gated Recurrent Unit (MH-GRU) for examining the weather patterns. From the weather patterns, the first level recommendation are generated using Fennec Fox Optimization (F2O) algorithm. Once the first level recommendations are generated, the input values are adjusted using normalization technique. From the adjusted values, the intelligent crop yield prediction is enabled by combining Deep Learning (DL) and transformer named Adaptive Hybrid Crop Yield Prediction Network (AHPCYPN). Finally, with the same F2O algorithm, we generate second level recommendation at the cloud layer. The proposed TLA-DSS showcases the superior performance in terms of performance metrics with the existing works and algorithms by offering scalable and robust solution for modern smart agriculture.