Single- and Multi-vehicle Safety Analysis for Internet Rental Bicycles Using Macro-Level Crash Prediction Models

Bicycle crashes are easy to cause casualties, therefore it is necessary to pay attention to bicycle traffic safety. In this study, 159 census tracts within Beijing Sixth Ring Road are utilized. Demographic features, area, location, road network features, and shared bicycle crash data are collected for the research units. Macro-level shared bicycle crash prediction models are developed for total, single- and multi-vehicle crashes separately using negative binomial models. The results show that 1) area and intersection density are positively associated with total shared bicycle crash frequency, while the percentage of population over 64 years old and the percentage of branches in the road network are negatively correlated with total shared bicycle crash frequency; 2) the increase of intersection density has a greater impact on single-vehicle crashes compared to multi-vehicle shared bicycle crashes; 3) higher population density is associated with higher multi-vehicle shared bicycle crash frequency, while it is not associated with single-vehicle crashes. This research provides a theoretical basis for road network safety planning in consideration of cyclists.