The effect of meteorology and air quality to the COVID-19 cases in Malaysia: a multivariate deep learning approach

Abstract

In October 2022, the World Health Organization (WHO) reported that over six hundred million people globally had been infected by the COVID-19 pandemic, leading to six million deaths. Malaysia, like many other countries, has experienced significant economic and societal impacts due to COVID-19. Previous research has identified meteorological conditions and air quality as critical factors influencing the spread of infectious diseases like influenza. In this study, we explore the impact of meteorological and air quality factors on COVID-19 case numbers in Malaysia, focusing on a case study in the Hulu Langat district of Selangor state, utilizing a deep learning approach. Our model, which employs a neural network architecture incorporating both Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN), was trained using multivariate time-series data. This data included meteorological and air quality metrics from the Department of Environment, Malaysia, and COVID-19 case data collected from the Hulu Langat Health Office. We prepared three datasets for predictive modeling: one combining all features, one including only meteorological data, and another with only air quality data. Our results indicate that the CNN model outperformed the LSTM model in terms of prediction accuracy. Furthermore, the dataset incorporating all features resulted in the lowest prediction error, compared to datasets with only meteorological or air quality features. Feature importance analysis showed that air quality factors were the most significant predictors, suggesting that air quality has a greater impact on COVID-19 case numbers than meteorological factors.