Agriculture Climate & Environmental AI
This project presents an explainable deep learning framework for rainfall prediction and agricultural risk assessment in Bangladesh. The proposed approach combines Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and an Attention mechanism to learn complex temporal patterns from historical temperature and rainfall data. The framework performs rainfall forecasting, classifies predicted rainfall into risk categories, and maps these predictions to potential agricultural impacts such as drought, flooding, and extreme flooding. SHAP-based explainability is incorporated to identify the most influential weather features behind model predictions. Seasonal analysis and a 12-month forecasting component further demonstrate how the framework can support climate-aware agricultural planning and decision-making.
More work
