Federated Learning-based Framework for Rainfall Prediction in Smart Agriculture
Bisma Mashooq, Yashwant Prasad Singh, Sheikh Imroza Manzoor · Procedia Computer Science · 2025
The Internet of Things (IoT), which powers smart agriculture, has a profoundly positive influence on people’s lives. Smart agriculture transforms farming techniques through soil quality detection, weather monitoring, flood prediction, rainfall prediction etc. Farmers can correctly monitor soil conditions, optimize irrigation and fertilizer management, and make wise decisions based on current weather information by using IoT devices and sensors. In smart agriculture, rainfall forecasting is essential because it helps farmers decide how to best use their resources and cultivate their crops. Traditional techniques for predicting rainfall heavily rely on centralized models that involve gathering and processing huge amounts of data from multiple meteorological sensors. However, these strategies frequently have problems including limited scalability and worries about data privacy. In our study, a novel federated learning approach for local model training using Artificial Neural Networks (ANN) is proposed for the prediction of rainfall in smart agriculture. Federated learning permits regional weather sensors to cooperatively train a common model while maintaining sensitive data’s confidentiality, ensuring data privacy and scalability. This paper provides a thorough explanation of the strategy, its implementation, and the experimental findings. With an accuracy rate of 84%, the suggested method for predicting rainfall events has proven to be useful.