Privacy-Preserving Crop Disease Prediction using Federated Meta-Learning and Climate Intelligence
The growing diversity of climate conditions and the scale effects of agricultural data pose serious problems for the accuracy and scalability of crop disease prediction. The drawbacks of traditional centralized machine learning methods include limitations caused by data privacy, low generalization, and high communication overheads. To overcome these challenges, in this study, we propose a new federated meta-learning with climate-driven personalization (FML-CDP) system that incorporates federated learning, meta-learning, and climate-sensitive modelling in a single architecture. The suggested system allows decentralized and distributed training to several farms without losing data privacy and addresses non-identically distributed (non-IID) data. The framework is more accurate in terms of predictions and more context-aware by using multimodal inputs, such as leaf images, IoT sensor data, and climate variables. The meta-learning aspect allows for quick local farm adjustments and individual predictions. The proposed model outperformed the baseline approach, achieving an accuracy of 97.3% and an F1-score of 0.96, surpassing the performance of centralised convolutional neural networks (CNN), FedAvg, and meta-learning models. Another benefit of climate intelligence is that it enhances resilience and flexibility of the system. The proposed framework offers privacy-saving, scalable, and efficient next-generation smart agricultural systems.