Logo Lanfrica

patience-2510789/maize_yield_prediction

Domain:

agriculture

Record type:

model
Creator:
pat
Host:
ML model to predict maize harvest in Uasin Gishu County, Kenya # 🌽 Maize Yield Prediction β€” Uasin Gishu County, Kenya A machine learning web application that predicts maize harvest yields in Uasin Gishu County using historical climate, satellite, soil, and seed variety data. πŸ”— **Live App:** maizeyieldprediction-kujsee… --- ## πŸ“Œ Project Summary | Item | Detail | |------|--------| | **Location** | Uasin Gishu County, Kenya | | **Target** | Maize Yield (Tonnes/Hectare) | | **Data Range** | 2005–2024 (20 years) | | **Model** | Linear Regression (RΒ² = 0.76, MAE = Β±0.30 t/ha) | | **Institution** | KCA University | --- ## πŸ—‚ Project Structure ``` maize-prediction/ β”œβ”€β”€ app.py # Main Streamlit app (Predictor) β”œβ”€β”€ requirements.txt # Python dependencies β”œβ”€β”€ pages/ β”‚ β”œβ”€β”€ 01_dashboard.py # Charts & insights dashboard β”‚ └── 02_methodology.py # Full methodology & documentation β”œβ”€β”€ notebooks/ β”‚ β”œβ”€β”€ 01_weather_download.ipynb # NASA POWER weather data β”‚ β”œβ”€β”€ 02_soil_data.ipynb # ISRIC soil data β”‚ β”œβ”€β”€ 03_ndvi_satellite.ipynb # MODIS NDVI satellite data β”‚ β”œβ”€β”€ 04_yield_and_seeds.ipynb # Yield & seed variety data β”‚ β”œβ”€β”€ 05_master_merge.ipynb # Merge all datasets β”‚ └── 06_modelling.ipynb # Model training & evaluation β”œβ”€β”€ data/ β”‚ β”œβ”€β”€ raw/ # Raw downloaded files β”‚ └── processed/ # Cleaned, merged datasets └── outputs/ # Charts and saved model ``` --- ## πŸ“‚ Data Sources | Data Type | Source | Years | |-----------|--------|-------| | Weather & Climate | NASA POWER | 2005–2024 | | Satellite NDVI | NASA MODIS via ORNL | 2005–2024 | | Soil Properties | ISRIC SoilGrids | Static | | Maize Yield | Uasin Gishu County Agriculture Office | 2012–2023 | | Seed Varieties | KEPHIS + CIMMYT | 2005–2024 | --- ## πŸ€– Model Performance | Model | MAE | RMSE | RΒ² | |-------|-----|------|----| | **Linear Regression** βœ… | **0.300** | **0.364** | **0.763** …