Transforming Air Quality Index Prediction Using Machine Learning: Key Insights from the Taj Trapezium Zone (TTZ)

Abstract

The exponential growth of air pollution is alarming for today's modern society. The Air Quality Index (AQI) is the main index used to gauge the severity of air pollution. In light of the recent Commission for Air Quality Management (CAQM) mandate for three days' advance upgrading of the Graded Response Action Plan (GRAP) stage, this research focuses on predicting the AQI to minimize negative economic losses through the correct adoption of the GRAP stage. The current research paper aims to predict the AQI in the Taj Trapezium Zone (TTZ) using machine learning algorithms. In this study, the main air contaminants that affect air quality—PM2.5, PM10, CO, NO2, NH3, NOx, O3, SO2, benzene, and toluene—along with meteorological factors such as temperature, humidity, wind speed, wind direction, and pressure, are used for data analysis and AQI prediction using machine learning algorithms. The study also aims to identify the most dominant pollutants in the TTZ area. The data source is a real-time dataset from air quality monitoring stations operated by the Central Pollution Control Board (CPCB) in Agra, one of the key cities within the TTZ area. This study uses four machine learning algorithms: AdaBoost, XGBoost, CatBoost, and LightGBM to calculate and compare AQI prediction accuracy. Four statistical performance metrics are used, namely: R2 score, Root Mean Square Error (RMSE), Mean Square Error (MSE), and Mean Absolute Error (MAE). The findings show that the XGBoost algorithm forecasts the AQI with the highest accuracy and best performance in the TTZ region. The study also found that PM10 is the most dominant pollutant in the TTZ area. This indicates that more stringent control measures are needed to curb PM10 pollution and improve the AQI. The use of these predicted AQI values may help CAQM in real-time GRAP stage decision making.