Apppublic

ZeniTsuIsSimP/npk-crop-recommendation

sourceHugging Facemitupdated 6mo agoView on Hugging Face

0likes

App README

🌾 NPK Crop Intelligence — MLOps Edition

![CI Pipeline](https://github.com/ZeniTsuisSimp/NPK-LIVE/actions/workflows/ci.yml) ![Hugging Face Spaces](https://huggingface.co/spaces/ZeniTsuisSimp/npk-crop-recommendation)

Live Project: https://huggingface.co/spaces/ZeniTsuisSimp/npk-crop-recommendation

Smart soil analysis and crop recommendation system powered by Machine Learning, built with MLOps best practices — DVC, MLflow, Docker, and CI/CD.

🏗️ Architecture

mermaid

graph LR
    A[📊 Dataset] -->|DVC tracked| B[🔧 Preprocessing]
    B --> C[🤖 Training]
    C -->|MLflow logged| D[📈 Evaluation]
    D --> E[✅ Model .pkl]
    E --> F[🌐 Streamlit App]
    F -->|Docker| G[☁️ Deployment]

    subgraph "CI/CD"
        H[GitHub Push] --> I[Lint + Test]
        I --> J[Train Pipeline]
        J --> K[Docker Build]
        K --> L[Deploy]
    end

📂 Project Structure

NPK-LIVE/
├── .github/workflows/      # CI/CD pipelines
│   ├── ci.yml               # Lint, test, train on push/PR
│   └── cd.yml               # Docker build + deploy on main
├── src/                     # Modular ML pipeline
│   ├── data_preprocessing.py
│   ├── train.py             # MLflow-integrated training
│   └── evaluate.py          # Metrics generation
├── app/                     # Streamlit application
│   └── npk_crop_recommendation_app.py
├── tests/                   # Unit tests
│   └── test_pipeline.py
├── data/                    # Raw dataset
│   └── Crop_recommendation.csv
├── models/                  # Trained model artifacts
│   └── npk_crop_model.pkl
├── notebooks/               # Jupyter notebooks
├── reports/                 # Auto-generated metrics
├── params.yaml              # Hyperparameter config
├── dvc.yaml                 # DVC pipeline stages
├── Dockerfile               # Container config
├── docker-compose.yml       # Docker Compose
├── requirements.txt         # Production deps
└── requirements-dev.txt     # Dev/test deps

🚀 Quick Start

1. Clone & Install

bash

git clone https://github.com/ZeniTsuisSimp/NPK-LIVE.git
cd NPK-LIVE
pip install -r requirements.txt
pip install -r requirements-dev.txt
pip install -e .

2. Run the ML Pipeline

bash

# Step-by-step
python -m src.data_preprocessing
python -m src.train
python -m src.evaluate

# Or use DVC
dvc repro

3. Launch the App

bash

streamlit run app/npk_crop_recommendation_app.py

4. Run with Docker

bash

docker-compose up --build
# Visit http://localhost:8501

🔬 MLOps Stack

Tool	Purpose
MLflow	Experiment tracking (params, metrics, model artifacts)
DVC	Data versioning & reproducible pipelines
Docker	Containerized deployment
GitHub Actions	CI (lint + test) and CD (Docker + deploy)
pytest	Unit testing
flake8	Code linting

View MLflow Experiments

bash

mlflow ui
# Visit http://localhost:5000

Run DVC Pipeline

bash

dvc repro           # Run full pipeline
dvc metrics show    # View latest metrics
dvc dag             # Visualize pipeline DAG

🧪 Testing

bash

pytest tests/ -v

🔄 CI/CD Pipeline

CI (on every push/PR)

1.Lint — flake8 syntax & style checks
2.Test — pytest unit tests
3.Train — Full preprocessing → training → evaluation
4.Verify — Model and metrics artifacts exist

GitHub Secrets Required

—HF_USERNAME — Hugging Face username
—HF_TOKEN — Hugging Face Write Token

📊 App Features

—🌾 Crop Prediction — ML-based crop recommendation from soil NPK values
—🧪 NPK Additions — Calculate nutrient amendments for target crops
—🔄 Crop Rotation Advisor — Science-based next-crop suggestions
—💚 Soil Health Score — Composite 0-100 health rating with radar chart
—📅 Seasonal Calendar — Indian Kharif/Rabi/Zaid crop calendar
—📈 NPK History — Track soil readings over time

📋 Model Details

Parameter	Value
Algorithm	Random Forest Classifier
Features	N, P, K (mg/kg)
Crops	Rice, Wheat, Corn, Barley, Soybean, Cotton, Sugarcane, Tomato, Potato, Onion
Config	`params.yaml`

Built with ❤️ using Streamlit, scikit-learn, MLflow, and DVC