Evaluation of machine learning in predicting air quality index / Abdullah Sani Abdul Rahman, Aizal Yusrina Idris and Suhaimi Abdul Rahman

Environmental pollution poses significant health risks, and Malaysia is facing a critical air pollution issue due to the rapid growth of urbanization and industrialization. The Air Quality Index (AQI) is a standard measure of air pollution, and machine learning methods have shown promise in accurately predicting AQI levels. However, there is limited research on the application of intelligent approaches to predict AQI in Malaysia. This research investigates the impact of various AQI components, including Particulate Matter 2.5 (PM2.5), Nitrogen Dioxide (NO2), Carbon Monoxide (CO), and Ozone (O3), using 125 random locations across Malaysia, ranging from the north to the southern regions. Three machine learning algorithms, namely Generalized Linear Model, Decision Tree and Support Vector Machine are used in this research. The results show that PM2.5 has the most significant impact on AQI levels among all components analyzed, and all selected machine learning algorithms exhibit high prediction accuracy, with R^ above 90% and low prediction errors (less than 2 MAE and RMSE). This research provides essential insights into predicting AQI levels using machine learning approaches and highlights the critical role of PM2.5 in determining AQI levels in Malaysia. The findings can aid authorities in obtaining rapid and accurate information to effectively manage air pollution in the country.