Machine Learning model trained on India-wide air quality data
Trained on 841,812 rows of hourly AQI data from 29 Indian cities. Best model highlighted in blue.
The prediction uses a hybrid ensemble:
Final AQI = (0.8 × LightGBM) + (0.2 × XGBoost)
Both models were fine-tuned separately using Special_Pipeline.py achieving ~80% accuracy on a no-pollutant feature set, and even higher when pollutant data is available.
ℹ️ Fill in the details below. Lag and rolling window features are automatically approximated from the values you enter — you don't need to calculate them manually.