A Data-Centric Approach to Weather Prediction: IoT and Machine Learning on Drone Platforms

The Internet of Things (IoT) is increasingly influential in transforming various sectors, including agriculture, by providing cost-effective sensors and easily deployable infrastructure. This adoption is revolutionizing the management of agricultural activities, shifting from traditional practices to data-centric and automated decision-making processes. Machine learning (ML) algorithms are integral to this transformation, particularly in predicting and forecasting weather conditions critical for agriculture. However, ML algorithms’ high computational and memory demands pose significant challenges for small IoT devices. The Emergence of Edge Devices with High Computation Capabilities may be a feasible solution. This paper proposes a drone-based weather prediction system designed to automate agriculture’s weather prediction and decision-making process. The DL/ML model may be trained on historical weather at high-computation facilities, and this pre-trained model is deployed on the drone-mounted edge node. The edge node-based DL/ML pre-trained model, integrated with IoT sensors, processes real-time weather data to make informed actuation decisions. Our experimental results, focused on rainfall prediction, show that the system achieves high accuracy, with logistic regression performing best on the Bikaner dataset and XGBoost excelling on the Australian data. The precise weather predictions enabled by our system lead to more accurate agricultural decisions, thereby optimizing the automation of agricultural processes.