🌱 AgroIntel Crop Advisor

A Streamlit application that recommends the optimal crop for a plot of land based on soil nutrients (N, P, K, pH) and environmental conditions (temperature, humidity, rainfall). It trains and compares classic machine learning models (Random Forest, KNN, SVM) on the Crop Recommendation dataset and exposes the results through an interactive UI for data exploration, model evaluation, and live predictions.

---

Features

The full application (simple_app.py) is organized into the following pages:

Page	What it does
Home	Dataset overview, summary statistics, and crop-distribution chart.
Data Analysis	Feature distributions, correlation heatmap, and soil-nutrient breakdown by crop.
Model Performance	Accuracy/precision/recall/F1 for Random Forest, KNN, and SVM, plus feature importance.
Prediction Tool	Enter soil/environmental values and get a recommended crop with suitability scores.
Feature Engineering	Polynomial features, ratio features, PCA, and automated feature selection.
Clustering Analysis	K-Means/DBSCAN clustering, optimal-k analysis, and cluster profiling.
Advanced Analysis	Model comparison, confusion-matrix analysis, ROC/PR curves, and learning/validation curves.

A second, simplified entry point (app.py) implements the Home, Data Analysis, Model Performance, and Prediction Tool pages as a smaller, modular reference (its Advanced Analysis, Advanced Modeling, and Seasonal Analysis pages are present in the UI but currently stubbed out).

---

Tech Stack

• Language: Python 3.11
• UI: Streamlit
• ML: scikit-learn (Random Forest, KNN, SVM, K-Means, DBSCAN, PCA)
• Data: pandas, numpy
• Visualization: Plotly, Matplotlib, Seaborn
• Model persistence: joblib
• Packaging: uv (pyproject.toml + uv.lock)
• Containerization: Docker / Docker Compose

---

Project Structure

.
├── app.py                   # Modular entry point (imports from src/)
├── simple_app.py            # Full-featured, self-contained entry point
├── src/                     # Shared analysis/modeling modules used by app.py
│   ├── data_preprocessing.py
│   ├── exploratory_analysis.py
│   ├── modeling.py
│   ├── utils.py
│   ├── feature_engineering.py
│   ├── clustering.py
│   ├── advanced_modeling.py
│   └── seasonal_analysis.py
├── data/
│   └── Crop_recommendation.csv
├── models/                  # Trained model artifacts (.joblib), created on first run
├── assets/                  # App icon and screenshots
├── .streamlit/config.toml   # Streamlit server config (headless, port 5000)
├── Dockerfile
├── docker-compose.yml
├── pyproject.toml / uv.lock # Python dependencies
└── .replit / replit.nix     # Replit deployment config

---

Running with Docker (recommended)

This spins up the full application — dependencies, data, and pretrained models — with a single command.

Requirements: Docker and Docker Compose.

git clone <repo-url>
cd AgroIntel_CropAdvisor
docker compose up --build

Then open http://localhost:5000 in your browser.

• The container runs simple_app.py by default.
• The models/ directory is mounted into the container, so any model retraining persists back to your host instead of being lost when the container stops.
• Stop the app with docker compose down.

---

Running locally (without Docker)

Requirements: Python 3.11+ and uv (or pip).

git clone <repo-url>
cd AgroIntel_CropAdvisor

# Using uv (matches the committed lockfile)
uv run streamlit run simple_app.py

# ...or with pip
pip install -e .   # or: pip install joblib matplotlib numpy pandas plotly scikit-learn seaborn streamlit
streamlit run simple_app.py

Then open http://localhost:5000 (configured in .streamlit/config.toml).

To run the modular version instead:

streamlit run app.py

---

Dataset

data/Crop_recommendation.csv contains 2,200 samples across 22 crop labels, with the following features:

Feature	Description
N	Nitrogen content in soil (kg/ha)
P	Phosphorus content in soil (kg/ha)
K	Potassium content in soil (kg/ha)
temperature	Temperature (°C)
humidity	Relative humidity (%)
ph	Soil pH value
rainfall	Rainfall (mm)
label	Target crop

---

Upcoming Enhancements

• Pest risk alerts based on environmental and historical data.
• Irrigation planning recommendations tailored to crop type and soil moisture.
• Market-driven crop selection based on demand trends.

---

License

No license has been specified for this project.