Air quality forecasting using a modified statistical approach: Combining statistical and machine learning methods

Abstract

Accurate prediction of air quality in urban areas is critical due to the increasing impact of pollution on public health and environmental sustainability. The purpose of this study is to develop an enhanced forecasting model for urban air quality using a hybrid approach. The methodology integrates statistical techniques, namely least absolute shrinkage and selection operator (LASSO), ridge regression, and elastic net, with machine learning models such as random forest (RF), k-nearest neighbor regression (KNN), and extreme gradient boosting (XGBoost). A novel ensemble model combining elastic net and RF is proposed to improve predictive accuracy. The approach was validated using a comprehensive air quality dataset collected from multiple Indian cities between 2015 and 2020. India was chosen because it is one of the largest countries in Asia. The findings indicate that the hybrid model outperforms traditional statistical and machine learning models in terms of predictive performance, as assessed by robust goodness-of-fit metrics. In conclusion, the proposed method provides a powerful and reliable tool for predicting air quality and thus achieving environmental sustainability and meeting the basic needs of humans. Practical implications of this work include its potential use by policymakers and environmental agencies for proactive pollution management and public health planning.