citywise-pm25-forecast

An AI-powered Streamlit web application to forecast PM2.5 air pollution levels for any Indian city using historical environmental data.

🚀 Features

• 📊 City-wise PM2.5 concentration prediction for next 3 months
• 🧠 Trained on XGBoost Regression Model
• 🏙️ Built using Indian air quality & meteorological datasets
• 📉 Auto-reshaping of historical wide-format data to time-series format
• 📦 Real-time visualization & download of forecast data
• 💡 Streamlit UI for easy use and deployment

📌 How It Works

1. Historical air quality data is loaded, cleaned, and reshaped (wide → long format)
2. Features like year, month, SO2, NO2, PM10, and previous PM2.5 values are used
3. XGBoost Regression is trained to forecast PM2.5 for the next 3 months
4. User selects city in Streamlit UI and gets predictions & plots

---

🧪 Model Details

• Model: XGBoostRegressor
• Target: PM2.5 lagged value
• Features: SO2, NO2, PM10, month, year, lagged PM2.5
• Evaluation: MAE, RMSE on validation set

---

📦 Installation

# Clone the repo
git clone https://github.com/Shraddha-DSA/citywise-pm25-forecast.git
cd citywise-pm25-forecast

# Install dependencies
pip install -r requirements.txt

# Run the app
streamlit run streamlit_app.py