Sarvjais12/Pollution_Anomaly_Detector

📉 Delhi Air Quality Forecaster (PM2.5)

🌍 About This Project

This application predicts future air pollution levels (PM2.5) in New Delhi, India. Delhi faces a severe air quality crisis every winter due to a combination of weather patterns, crop burning, and vehicular emissions. This tool aims to forecast these hazardous "smog spikes" months in advance, demonstrating how machine learning can aid in proactive crisis management.

🧠 How It Works

The model is built on historical daily air quality data and uses SARIMA (Seasonal AutoRegressive Integrated Moving Average).

• —Why SARIMA? Unlike standard models, SARIMA specifically understands "seasonality"—the repeating 12-month cycle of Delhi's pollution.
• —Training Data: The model learned from years of historical PM2.5 readings, identifying that pollution consistently peaks in winter (Nov-Jan) and drops in summer.

🚀 How to Use

1. 1.Adjust the Slider: Select how many months into the future you want to forecast (e.g., 6 months, 12 months).
2. 2.View the Graph: 🔵 Blue Line:* The model's understanding of recent history.
3. 3.🔴 Red Dashed Line: The predicted future pollution levels.
4. 4.Analyze: Look for the sharp spikes in the red line—these indicate predicted hazardous air quality periods.

⚠️ Disclaimer

This is a machine learning project for educational and research purposes. While the model captures seasonal trends effectively (RMSE ~27.5), real-world air quality is also affected by unpredictable daily factors like sudden rain, wind speed, or policy changes.

Created by Sarvagya Jaiswal | Powered by Scikit-Learn, Statsmodels, and Gradio.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference