AI-IoT driven system for agricultural pest outbreak risk prediction

Abstract Invasive pests pose a significant threat to agricultural production, particularly maize crops, with severe implications for food security. Timely detection of pest development stages and accurate prediction of outbreak risks are essential for effective management. This study introduces a hybrid model combining Explainable Artificial Intelligence (XAI), a lightweight Convolutional Neural Network (CNN), and Fuzzy Logic (FL) for Fall Armyworm (FAW) detection and weather-based risk prediction. The model uses Tiny-MobileNet-SE for image classification, Grad-CAM for interpretability, and FL inference based on environmental parameters. Tiny-MobileNet-SE achieved 98.6% accuracy, 98.5% F1-score, 98.6% recall, a compact size of 0.72 MB, and 80 ms latency on Raspberry Pi 5, outperforming state-of-the-art lightweight models including EfficientNetB0, SqueezeNet, MobileNet-v2, MobileNet-v3, and ShuffleNet. The proposed system delivers a power-efficient, scalable, and user-friendly solution for precision agriculture, providing actionable insights for pest management and supporting sustainable crop protection strategies.