🌾 AgriAdvisor

An AI-powered agricultural advisory web application that helps Egyptian farmers make data-driven decisions about crop selection, yield estimation, and profit forecasting.

---

📋 Table of Contents

• Overview
• Features
• Tech Stack
• Project Structure
• Getting Started
• API Reference
• ML Models
• Screenshots

---

Overview

AgriAdvisor is a graduation project that combines machine learning with a web interface to assist farmers across Egypt. Users can input soil type, region, season, and land area to receive intelligent recommendations powered by trained ML models.

The backend is built with FastAPI and serves three prediction endpoints. The frontend consists of five HTML pages — a landing page, three tool pages, and a crop guide.

---

Features

Page	Feature	ML Model
Page 2 — Crop Recommender	Recommends the best crop based on soil, region, and season	LightGBM / XGBoost classifier
Page 3 — Yield Estimator	Estimates yield in tons per feddan	SVR / Random Forest regressor
Page 4 — Profit Calculator	Forecasts expected profit in EGP	XGBoost regressor (R²=0.93)
Page 5 — Crop Guide	Static guide for 8 major Egyptian crops	—

Supported crops: Wheat, Rice, Maize, Cotton, Sugarcane, Tomato, Potato, Fava Beans

---

Tech Stack

Backend

• Python 3.x
• FastAPI
• Uvicorn
• scikit-learn, XGBoost, LightGBM
• joblib

Frontend

• HTML5 / CSS3 / JavaScript (Vanilla)
• Fetch API for backend communication

---

Project Structure

Website_Project/
│
├── main.py                          ← FastAPI entry point
│
├── routers/
│   ├── crop.py                      ← POST /crop/predict
│   ├── yield_.py                    ← POST /yield/predict
│   └── profit.py                    ← POST /profit/predict
│
├── ml_models/
│   ├── crop.py                      ← Crop inference logic + CLIMATE_DEFAULTS
│   ├── yield_.py                    ← Yield inference logic
│   └── profit.py                    ← Profit inference logic
│
├── Page One/
│   └── AgriAdvisor-index.html       ← Landing page
│
├── Page Two/
│   ├── page2.html                   ← Crop Recommender UI
│   ├── Images/                      ← Crop photos (8 images)
│   └── models/
│       ├── crop_model.pkl           ← Trained classifier
│       └── crop_recommendation.csv  ← Training dataset (7,000 rows)
│
├── Page Three/
│   ├── AgriAdvisor-page3-yield.html ← Yield Estimator UI
│   └── Model/
│       ├── yield_model.pkl          ← Trained regressor
│       └── egypt_yield_8400.csv     ← Training dataset (8,400 rows)
│
├── Page Four/
│   ├── AgriAdvisor-page4-profit.html ← Profit Calculator UI
│   └── Model/
│       ├── profit_model.pkl          ← Active model (XGBoost)
│       └── profit_estimation_10k.csv ← Training dataset (10,400 rows)
│
└── Page Five/
    ├── AgriAdvisor-page5-guide.html  ← Crop Guide
    └── images/                       ← 32 crop photos (4 per crop)

---

Getting Started

Prerequisites

pip install fastapi uvicorn scikit-learn xgboost lightgbm joblib

Run the Server

cd Website_Project
uvicorn main:app --reload

The API will be available at http://localhost:8000

Interactive API docs: http://localhost:8000/docs

Open the Frontend

Open Page One/AgriAdvisor-index.html in your browser and navigate from there.

Note: The frontend connects to http://localhost:8000 by default. Make sure the FastAPI server is running before using any of the tool pages.

---

API Reference

POST /crop/predict

Recommends the best crop for given conditions.

{
  "region": "Delta",
  "season": "Winter",
  "soil": "Clay"
}

POST /yield/predict

Estimates yield per feddan and total yield.

{
  "crop": "Wheat",
  "season": "Winter",
  "region": "Delta",
  "soil": "Clay",
  "area_feddans": 5
}

Response includes: yield_per_feddan, total_yield, soil_efficiency, bar_pct

POST /profit/predict

Calculates expected profit.

{
  "crop": "Wheat",
  "season": "Winter",
  "region": "Delta",
  "soil": "Clay",
  "area_feddans": 5
}

All endpoints return 400 for invalid crop–season or crop–soil combinations.

---

ML Models

Crop Recommender (Page 2)

• Algorithm: LightGBM / XGBoost (classifier)
• Input features: 11 features built from 3 user inputs via CLIMATE_DEFAULTS lookup table
• Dataset: 7,000 rows — crop_recommendation.csv
• Training notebooks: SVM.ipynb, LightGBM.ipynb, XGBoost.ipynb

Yield Estimator (Page 3)

• Algorithm: SVR / Random Forest (regressor)
• Dataset: 8,400 rows — egypt_yield_8400.csv
• Training notebooks: SVR.ipynb, RandomForest.ipynb

Profit Calculator (Page 4)

• Algorithm: XGBoost regressor — R² = 0.93
• Dataset: 10,400 rows — profit_estimation_10k.csv
• Training notebooks: LightGBM.ipynb, XGBoost.ipynb

All models fall back to a demo/lookup-table mode if the .pkl file is not found, so the app stays functional for testing.

The trained model files are included in the repository and loaded automatically on startup.

---

Screenshots

Landing Page

Crop Recommender

Yield Estimator

Profit Calculator

---

.gitignore Recommendations

*.pkl
__pycache__/
*.pyc

---

Authors

Built as a graduation project by Ramez Malak — Arab Open University (AOU).

---

License

This project is for academic purposes only.