title	emoji	colorFrom	colorTo	sdk	sdk_version	app_file	pinned
Kaduna Agricultural Forecasting	🌾	green	yellow	streamlit	1.25.0	app.py	false

Agricultural Production Forecasting System for Kaduna State

This system provides advanced time-series forecasting for crop production in Kaduna State to support food security planning and policy decisions. It implements a three-tier modeling strategy using Statistical (ARIMA/SARIMAX), Machine Learning (XGBoost), and Deep Learning (LSTM/GRU) models.

Features

• Multi-Model Forecasting: Compare performance of ARIMA, SARIMAX, XGBoost, LSTM, and GRU models.
• Automated Feature Engineering: Generates yield, lag features, rolling statistics, and trend indicators.
• Policy Recommendations: Automatically detects shortage/oversupply risks and suggests policy actions.
• Interactive Visualizations: Generates Plotly charts for forecasts and model comparisons.
• Comprehensive Evaluation: Uses RMSE, MAE, MAPE, and R² to evaluate model accuracy.

Installation

1. Clone the repository.
2. Install dependencies:

   pip install -r requirements.txt

Usage

Run Full Pipeline (Train & Forecast)

To train all models and generate a 3-year forecast for Maize:

python main.py --crop Maize --model all --train --forecast-years 3

Command Line Arguments

• --data: Path to dataset CSV (default: Kaduna_State_Agricultural_Production_Data.csv)
• --crop: Crop name to forecast (e.g., "Maize", "Sugar cane", "Rice")
• --model: Model to use (arima, sarimax, xgboost, lstm, gru, or all)
• --train: Flag to retrain models
• --forecast-years: Number of years to forecast (default: 3)

Output

Results are saved in the outputs/ directory:

• outputs/models/: Saved trained models
• outputs/predictions/: Forecast CSV files
• outputs/visualizations/: Interactive plots (HTML) and static charts (PNG)
• outputs/: Policy briefs (TXT) and model comparison tables (CSV)

Deployment

Interactive Web Dashboard

To launch the user-friendly web interface:

1. Install Streamlit:

   pip install streamlit

2. Run the app:

   streamlit run app.py

The dashboard allows you to:

• Select any crop and model
• Adjust forecast horizons
• View interactive charts
• Download policy briefs

KadPred/
├── src/
│   ├── models/         # Model implementations (ARIMA, XGBoost, LSTM, etc.)
│   ├── data_loader.py  # Data loading and cleaning
│   ├── feature_engineering.py # Feature creation
│   ├── evaluation.py   # Metrics and comparison
│   ├── visualization.py # Plotting utilities
│   ├── predictor.py    # Forecasting pipeline
│   └── policy_advisor.py # Policy recommendation logic
├── outputs/            # Generated results
├── main.py             # Main execution script
└── requirements.txt    # Dependencies

Policy Advisor Logic

The system analyzes forecasts against a 5-year historical baseline:

• Shortage Risk: Production < 20% of baseline (Action: Release reserves, Import)
• Oversupply Risk: Production > 20% of baseline (Action: Storage, Export)
• Normal: Within ±10% of baseline (Action: Maintain support)