DefinePK

Abstract

Accurate forecasting of the Air Quality Index (AQI) is essential for proactive environmental management and public health advisories. This study investigates and compares the predictive capabilities of four supervised learning models—Random Forest, XGBoost, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU)—for AQI forecasting in Bhopal, Madhya Pradesh, using hourly pollutant and meteorological data from February 2025. The dataset, collected from CPCB monitoring stations, includes key pollutants (PM2.5, PM10, NO₂, SO₂, CO, O₃) and meteorological parameters (temperature, humidity, wind speed, pressure, and rainfall). All models were evaluated using MAE, RMSE, MAPE, and R² metrics. Results indicate that deep learning models, especially GRU, outperform traditional machine learning models, achieving an R² of 0.952 and RMSE of 14.41. Feature importance analysis highlights PM2.5 and PM10 as dominant contributors to AQI variations. This study underscores the potential of recurrent neural networks for short-term AQI forecasting and provides a foundation for developing real-time, location-specific environmental alert systems.