Predictive Analysis of Pune City’s Air Quality Index: A Machine Learning and Time Series Approach

The rise in pollution and global warming poses a growing threat to the environment and public health, with industries contributing to air pollution and affecting human health and ecological balance. The impact of bad air quality can be quite broad and affect different aspects of life. To address the concern, a good analysis for the air quality index is needed. This paper helps in monitoring the trends, analysis of the air quality indices for the particular areas of Pune city and helps predict the Air Quality Index (AQI). This study conducts an investigative analysis using various machine learning techniques and time series forecasting methods to predict AQI levels. Each model is trained to predict AQI values based on the pollutant concentrations. Subsequently, the predicted AQI values are categorized into six distinct levels according to a predefined scale. The experimental evaluation demonstrates that the CatBoostRegressor model outperforms the others used here, boasting an RMSE as low as 4.73 and an adjusted R-squared value of 0.98. While the ARIMA model provides a decent forecast that aligns well with the actual data, its error rate is marginally lower than that of the CatBoostRegressor.