INJURY SEVERITY ANALYSIS OF ROAD TRAFFIC ACCIDENTS IN HONG KONG: A COMPARATIVE STUDY BEFORE AND DURING THE COVID-19 PANDEMIC

Road traffic accidents represent a substantial public health challenge in Hong Kong, resulting in significant daily casualties. Additionally, the outbreak of the COVID-19 pandemic has brought about unprecedented changes in the traffic volume, driver behavior, and sample selectivity, which could alter the transportation pattern and influence the injury severity of road traffic accidents. Therefore, understanding the impact of COVID-19 on transportation patterns is crucial for gaining insights into the implementation of road safety measures. Using a dataset collected from the Hong Kong Transport Department and the Hong Kong Police Force's Traffic Information System, this study employed the Random Forest and K-Nearest Neighbors to investigate impacts of multiple factors associated with crashes, drivers, vehicles, environment, and districts. Besides, feature importance was extracted and compared for examining the influence of contributing factors on severities of accidents before (2018-2019) and during the COVID-19 pandemic (2020-2022). The results indicate that the Random Forest algorithm demonstrated a robust prediction performance, achieving an average accuracy of 92.46%, a precision of 92.56%, a recall of 99.8%, and an F1-score of 96%. It was also found that the type of collision and vehicle movements had significant impacts in the pre-pandemic period, while the location of injury had a high contribution during the pandemic period. The insights presented in this paper have the potential to serve as valuable references to enhance road traffic safety in Hong Kong.