eben05-maker/harvestguard-ghana

AI-powered post-harvest loss prediction tool fot Ghanaian farmers # 🌽 HarvestGuard Ghana *AI-Powered Post-Harvest Loss Prediction for Ghanaian Farmers* > Know When to Sell. Know Where to Sell. Lose Less. ## About HarvestGuard Ghana is a machine learning web application that helps Ghanaian smallholder farmers predict post-harvest crop losses and find the best market to sell their produce across all 16 regions of Ghana. Built for the *Ghana AI Innovation Challenge 2026* — Agriculture Focus Area. ## Problem Ghana loses an estimated *$1.9 billion* annually to post-harvest losses. Farmers lack real-time information on: - How fast their crop is deteriorating - Which market offers the best price - How to store their produce properly ## Solution HarvestGuard Ghana uses a *Random Forest ML model* (R²=0.9873) trained on research-backed Ghanaian agricultural data to predict loss rates and provide actionable market intelligence. ## Features - Post-harvest loss prediction for Maize, Tomatoes, and Yam - Market price comparison across all 16 regions of Ghana - Risk alert system (High / Moderate / Low) - Loss progression chart over 60 days - Storage recommendations based on MoFA Ghana guidelines - Collective impact calculator ## ML Model - Algorithm: Random Forest Regressor (scikit-learn) - Training samples: 3,000 - R² Score: 0.9873 - Mean Absolute Error: 2.58% - Key feature: Storage method (82.9% importance) ## Data Sources - Ministry of Food & Agriculture Ghana (MoFA) SRID - Wongnaa et al. (2023) — Cogent Food & Agriculture - FAO / FAOSTAT — Ghana Agricultural Statistics - Selinawamucii.com (2025) — Ghana Retail Prices - Esoko Ghana Market Price Index (2024) - MoFA/NAFCO Guaranteed Farmgate Prices (Sep 2025) ## How to Run ```bash pip install -r requirements.txt python train_model.py streamlit run app.py