A Systematic and Comprehensive Study on Exploring Machine Learning and Deep Learning Techniques for Air Quality Index Prediction

Air pollution significantly impacts human health, ecosystems, and climate, necessitating accurate forecasting for effective mitigation strategies. This study utilizes daily air quality data from 2019 to late 2023 collected at a CPCB station in Vapi, India. Authors have predicted the Air Quality Index (AQI) using advanced machine learning and deep learning models. The dataset includes key pollutants such as PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">10</inf>, PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2.5</inf>, Ozone, SO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>, NO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>, and CO, addressing missing data challenges through robust imputation techniques, with the linear method proving most effective. Through hyperparameter tuning using Grid Search, we optimized model parameters for Random Forest, SVM, LSTM, GRU, BiLSTM, CNN-LSTM, and CNN-GRU, which were evaluated using R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, RMSE, and MAE metrics. CNN-GRU demonstrated superior performance for 15-day forecasts, while BiLSTM and LSTM excelled in 5-day forecasts. This framework combines traditional and deep learning approaches to provide accurate AQI predictions, enabling better public health protection, policymaking, and environmental management.