An intelligent decision-support system that recommends the most suitable crop based on soil, district, season, and water requirement — empowering farmers and agri-businesses to make informed, data-driven crop selection decisions.
"Sow data. Reap insights. Cultivate prosperity." 🌾
Agriculture faces challenges like fluctuating weather, soil degradation, and uncertain market demand. Our system leverages machine learning on real agri-market and soil data to:
- Recommend the best crop to cultivate in given conditions
- Optimize productivity, reduce risk, and maximize profitability
- Empower farmers with smart agri-decision tools
Tech Stack:
- Python (Pandas, Scikit-learn)
- Streamlit (for user-friendly web app)
- Machine Learning (Decision Trees)
The dataset used contains real-world agricultural market and soil parameters:
| Feature | Description |
|---|---|
| District | Name of district (e.g., Pune, Nashik) |
| APMC_Name | Agriculture Produce Market Committee Name |
| Commodity | Crop name (target variable) |
| Arrival_Quantity_Qtl | Quantity of crop arrived in market (Qtl) |
| Modal_Price_Rs_Qtl | Modal price (Rs/Qtl) |
| Soil_Type | Loam, Sandy etc. |
| Water_Requirement | Low, Medium etc. |
| Season | Kharif, Rabi |
| Average_Yield_Tons_per_Ha | Yield |
| Typical_Costs_Rs_per_Ha | Cost of production |
| Export_Potential | High, Medium, Low |
✅ Crop recommendation based on soil, water, and district ✅ Simple, interactive Streamlit web interface ✅ Easy-to-interpret Decision Tree model ✅ Export potential insights for market-driven decisions ✅ Scalable framework to integrate larger datasets
We trained a Decision Tree Classifier on encoded categorical features (District, Soil, Season, Water). The model predicts the Commodity most suitable under the given conditions.
- Data preprocessing and label encoding
- Feature selection: District, Soil_Type, Season, Water_Requirement
- Model training and evaluation
- Deployment via Streamlit app
This project demonstrates:
- Applied Machine Learning in Agriculture: Bridging the gap between agri-expertise and data science
- Feature Importance Analysis: Understanding key factors influencing crop suitability
- Decision Support System Design: Translating research into actionable tools for end-users
- Scalability Potential: Framework adaptable to integrate weather, market prices, and geospatial data in future iterations
# Clone the repo
git clone https://github.com/Atharvak29/Farm-Crop-Advisor.git
cd Farm-Crop-Advisor
# Install dependencies
pip install -r requirements.txt
# Run Streamlit app
streamlit run main.py| Model KeyFile | Purpose | Used In |
|---|---|---|
| incropmodel_crop_suitability.pkl | Crop suitability prediction | predict_crop_suitability() (inside recommendations) |
| yieldmodel_yield_forecast.pkl | Yield forecast | predict_yield() |
| pricemodel_price_forecast.pkl | Market price prediction | predict_market_price() |
| profitmodel_profit_estimation.pkl | Profit estimation | predict_profitability() |
| riskmodel_risk_assessment.pkl | Risk assessment | predict_risk() |
| Gap / Missing Item | Reason it’s needed | Suggestion |
|---|---|---|
| Suitability Score is random (not from model) | Currently assigning random suitability score np.random.randint(70,100) | Use actual model prediction probability (normalized as %) |
| Real time weather api missing | Currently using random Temerature, Rainfall, Humidity | Integrate weather api to get best predictions |
| Real District names insted of numbers | User can't understand what state they are choosing | Endcode the data properly or showcase a tables vs the encoded value for user reference |
| ETL Framework missing | Make a etl pipline that extracts data for real time data decision | Run a cronjob every data and fetch essesntial data from government website |
- Integrate real-time weather and market data
- Deploy advanced models (Random Forests, XGBoost)
- Enhance UI/UX and multilingual support
- Collaborate with agri-research bodies for larger datasets
- Implement mobile app
