🌍 AQI Prediction using Machine Learning

Predict the Air Quality Index (AQI) from real‑world pollutant data (CO, NO₂, SO₂, O₃, PM2.5, PM10) across multiple global cities using Python & ML.

---

---

✨ Features

• 📊 Exploratory Data Analysis – Visual pollutant distributions & AQI trends.
• 🧹 Data Cleaning – Null checks, duplicates, drop unused columns.
• 🧮 Feature Engineering – One‑hot encode cities; scale numeric features.
• 🤖 Machine Learning Models – Linear Regression baseline vs Random Forest ensemble.
• 📈 Model Evaluation – R², RMSE, MAE comparison.
• 🔮 Custom Prediction – Plug in new pollutant readings and estimate AQI.

---

🎯 Why This Matters

Poor air quality affects respiratory health, productivity, and urban planning. An ML model that estimates AQI from pollutant levels helps:

• Citizens track exposure risk.
• City agencies forecast alerts.
• Students learn regression modeling on environmental data.

---

📂 Dataset

Rows: 52,560 hourly records
Columns: City, CO, NO2, SO2, O3, PM2.5, PM10, AQI
Cities Covered: Brasilia, Cairo, Dubai, London, New York, Sydney
Use: Educational / learning project dataset (bundled locally in repo).

If you later host the dataset separately (e.g., Kaggle), update the link here.

---

🛠 Tech Stack

• Python (pandas, numpy)
• Visualization: matplotlib, seaborn
• Modeling: scikit-learn (LinearRegression, RandomForestRegressor, MinMaxScaler, metrics)
• Environment: Jupyter Notebook

---

🔍 Workflow

1. Load CSV → pandas.read_csv()
2. Inspect shape, dtypes, nulls
3. Drop Date (not modeled)
4. Encode City → one-hot columns
5. Split train/test
6. Scale features → MinMaxScaler
7. Train models:
   • Linear Regression (baseline)
   • Random Forest Regressor (ensemble)
8. Evaluate → R², RMSE, MAE
9. Predict on new samples

---

📊 Results

Model	R²	RMSE	MAE	Notes
Linear Regression	0.83	10.21	7.38	Baseline
Random Forest	0.86	9.37	6.33	✅ Best Model

(Metrics from notebook run; will vary by random seed.)

---

🚀 Quickstart

First create the repo on GitHub under your account prachi757 named aqi-prediction (Public). Then run the steps below.

1. Clone the repository

git clone https://github.com/prachi757/aqi-prediction.git
cd aqi-prediction

If you forked this repo instead: replace the URL with your fork (shown on GitHub after you click Fork).

2. (Optional) Create & activate a virtual environment

macOS / Linux

python -m venv .venv
source .venv/bin/activate

Windows PowerShell

python -m venv .venv
.\.venv\Scripts\activate

3. Install dependencies

pip install --upgrade pip
pip install -r requirements.txt

4. Launch the notebook

jupyter notebook Major_Project.ipynb

Run cells top→bottom.

---

🧪 Try Your Own Prediction

After running the notebook and training the Random Forest model:

# Example new pollutant reading (scaled automatically below)
# Order: CO, NO2, SO2, O3, PM2_5, PM10, Brasilia, Cairo, Dubai, London, New_York, Sydney
new_sample = [[0.7, 45.0, 12.0, 32.0, 58.0, 105, 0, 1, 0, 0, 0, 0]]

# IMPORTANT: Use the *same* scaler fitted on training data
new_sample_scaled = scaler.transform(new_sample)

pred = AQI_Regressor.predict(new_sample_scaled)
print(f"Predicted AQI: {pred[0]:.2f}")

---

📁 Project Structure

aqi-prediction/
│
├── Major_Project.ipynb        # Notebook: EDA + Modeling
├── Air_Quality_dataset.csv    # Dataset (hourly pollutant readings)
├── requirements.txt           # Environment + install instructions
└── README.md                  # You are here!

---

🗺 Roadmap (Future Ideas)

• Add Gradient Boosting / XGBoost
• Include time‑series features from Date (hour, month, season)
• Hyperparameter tuning (GridSearchCV)
• Streamlit mini‑app for live AQI prediction
• Feature importance + SHAP explainability

---

🙋 Contact

Prachi Garg
GitHub: prachi757
LinkedIn: Prachi Garg
Email: [email protected]

---

📜 License

Educational & portfolio use. Feel free to fork, learn, and extend—please credit the original author.

---

⭐ Like this project?

If it helped you, star the repo and share! 🙌