<p>The global impact of the COVID-19 pandemic is widely recognized as a significant concern, with human flow playing a crucial role in its propagation. Consequently, recent research has focused on identifying and analyzing factors that can effectively regulate human flow. However, among the multiple factors that are expected to have an effect, few studies have investigated those that are particularly associated with human flow during the COVID-19 pandemic. In addition, few studies have investigated how regional characteristics and the number of vaccinations for these factors affect human flow. Furthermore, increasing the number of verified cases in countries and regions with insufficient reports is important to generalize conclusions. Therefore, in this study, a group-level analysis was conducted for Narashino City, Chiba Prefecture, Japan, using a human flow prediction model based on machine learning. High-importance groups were subdivided by regional characteristics and the number of vaccinations, and visual and correlation analyses were conducted at the factor level. The findings indicated that tree-based models, especially LightGBM, performed better in terms of prediction. In addition, the cumulative number of vaccinated individuals and the number of newly infected individuals are likely explanatory factors for changes in human flow. The analyses suggested a tendency to move with respect to the number of newly infected individuals in Japan or Tokyo, rather than the number of new infections in the area where they lived when vaccination had not started. With the implementation of vaccination, attention to the number of newly infected individuals in their residential areas may increase. However, after the spread of vaccination, the perception of infection risk may decrease. These findings can contribute to the proposal of new measures for efficiently controlling human flows and determining when to mitigate or reinforce specific measures.</p>
