Coronavirus disease 2019 (COVID-19), an infectious disease caused by severe acute respiratory distress syndrome coronavirus-2, was first reported in December 2019 as pneumonia of unknown cause in Wuhan, China. On March 11, 2020, the World Health Organization (WHO) declared SARS-CoV-2 as a public health concern and pandemic. COVID-19 had infected more than 500 million people and caused 6.65 million deaths as of November 29, 2022 (World Health Organization et al., 2020). At the beginning of 2020, the disease rapidly spread worldwide, but the initial dynamics looked very different. The weekly increasing trend of the number of COVID-19 cases varied between countries. While some countries were able to slow down the growth and spread of COVID-19, others saw rapid and near-exponential growth (Dong et al., 2020). Hence, understanding the region-specific factors that influence outbreak occurrence is important in enhancing long-term defenses against ongoing pandemics. According to Ficetola and Rubolini (2021) predicting global infection risk is critical to the identification of the driving patterns of COVID-19 outbreaks. A study by Ilin et al. (2021) showed the effect of mobility restriction on the number of cases in a country. An increasing residential time would positively affect future infections. Their study estimated that a country’s cumulative infections will drop to 82.5% of what they would have otherwise had ten days later if a policy increasing residential time by 5% is implemented.

Besides mobility restrictions, vaccination has been shown to reduce the virus transmission more efficiently (Lin et al., 2020). According to WHO, as of November 29, 2022, almost 13 billion vaccine doses have been administered (Anon, 2023b). Countries should continue to strive to vaccinate at least 70% of their population, with programs for 100% of health workers and 100% of the most vulnerable groups, such as people aged over 60 years and those with compromised immune systems or managed diseases, with situational priority health (Anon, 2023c, Bell et al., 2023). Vaccines protect those who receive them and prevent the spread of disease throughout the population (Forni and Mantovani, 2021, Hayawi et al., 2021).

Some studies have associated temperature, humidity, season, and ultraviolet light, to the emergence of many infectious diseases, including COVID-19 (Carlson et al., 2020). A previous study by Shen et al. (2020) investigated the impact of social distancing to COVID-19 transmission. Studies by Hoogeveen et al., 2022, Muttaqien et al., 2022, Kuo and Fu, 2021 demonstrated that weather and human mobility could drive the infectious diseases. Furthermore, the study by Kubota et al. (2020) showed that the COVID-19 pandemic was driven by cross-border human mobility, climate conditions, and region-specific COVID-19 susceptibility. Their study implemented multiple linear regression and random forest (RF) to investigate the association among the drivers. They found that climate factor (temperature) has a minor effect on COVID-19. They also mentioned that the relative importance of the factors depends on the period. Then, different periods may result in different findings. In their study, the regression coefficients of the model significantly changed from nonsignificant to significant over the period of December 2019 to April 2020. In the beginning, the disease transmission mostly depended on the dispersibility of the virus. As the pandemic was progressing, the variables of climate, human mobility, and host susceptibility exhibited increasing importance in April 2020. After this time, the importance of human mobility and host decreased, whereas human population and climate factors increased.

Previous studies on COVID-19 drivers were mainly conducted during the start of the pandemic or used the data of the first year of the COVID-19 spread (Gupta et al., 2021, Kubota et al., 2020, Iwendi et al., 2020, Rath et al., 2020). Studies on COVID-19 using machine learning (ML) approaches, such as those conducted in Ogunjo et al., 2022, Karmokar et al., 2022, Chakraborti et al., 2021, Li et al., 2021b, have yielded favorable results in predicting factors associated with COVID-19 cases. The study (Ogunjo et al., 2022) used four different ML methods for virus infection prediction and found that temperature and relative humidity are substantial predictors of the number of COVID-19 cases. A study by Karmokar et al. (2022) also used ML to understand the characteristics of COVID-19 against weather conditions. ML regression models had also been shown to be effective in comprehensively identifying the critical factors responsible for the increase in the number of COVID-19 cases and mortalities (Chakraborti et al., 2021, Li et al., 2021b). Both studies (Chakraborti et al., 2021, Li et al., 2021b) investigated the beginning of the pandemic worldwide, except for some countries with unavailable data.

In this study, the potential factors affecting the spread of COVID-19 worldwide were assessed in a more extended period by analyzing the data of confirmed COVID-19 cases obtained from March 2020 to March 2022. As recent studies demonstrated that weather conditions, human mobility, and vaccinations affect the spread of COVID-19, this study explored the role of the three aforementioned variables in COVID-19 progression. Three ML models, namely RF, Extreme Gradient Boosting (XGBoost), and neural networks (NN) were applied because of their powerful methods and excellent performance to predict COVID-19 cases in previous studies (Gupta et al., 2021, Fang et al., 2022, Niazkar and Niazkar, 2020).