Stephen-Austine/All-derpin-data-challenge-2025

Hackathon Award - 3rd Place This project was developed during the All-Derpin Data Challenge 2025 and secured 3rd place, earning recognition for its innovative approach to agricultural climate forecasting in Malawi. # Malawi Forecasts - Crop Prediction System > **Hackathon Award - 3rd Place** > > This project was developed during the All-Derpin Data Challenge 2025 and secured **3rd place**, earning recognition for its innovative approach to agricultural climate forecasting in Malawi. --- ## Overview **Malawi Forecasts** is an intelligent early warning crop prediction system designed to help farmers and agricultural planners make data-driven decisions about crop selection. By leveraging advanced time-series forecasting (SARIMAX) and machine learning, the system predicts key climate variables—rainfall, vegetation health (NDVI), and land surface temperature—up to 5 years ahead. The system provides **crop recommendations** (Rice, Cassava, Maize) based on forecasted conditions, helping stakeholders plan for food security and agricultural resilience in changing climates. --- ## Hackathon Achievement | Achievement | Details | |-------------|---------| | **Position** | 3rd Place | | **Award** | All-Derpin Data Challenge 2025 | | **Team** | Stephen W. Austine, Andy E. Hadulo, George M. Rading | We are proud to have developed a solution that addresses real agricultural challenges in Malawi through data-driven insights. --- ## Key Features ### Climate Forecasting - **SARIMAX Modeling**: Predicts rainfall, NDVI, and temperature with 95% confidence intervals - **Multi-horizon forecasts**: National (3 years) and regional (5 years) predictions - **Auto-training**: Automatically trains and caches models for future use ### Crop Recommendations - **Intelligent analysis**: Recommends top 3 suitable crops based on forecasted conditions - **Least recommended**: Identifies crops to avoid under predicted conditions - **Reasoning**: Provides detailed explanations for each recommendation ### Model Evaluation - **Comprehensive metrics**: Accuracy, Precision, Recall, F1-Score - **Confusion matrix**: Visual assessment of model performance - **Historical validation**: Evaluates against ac …