Highlights We detect bicycle wrong way riding (WWR) behavior using bike-sharing data. Negative Binomial-based Additive Decision Trees are developed. The non-linear effects of built environment on WWR frequency are captured. WWR is unaffected by different times of day. Residence, company and transit facility are the most influential factors.
Abstract Bicycle wrong way riding (WWR) is a dangerous and often neglected behavior that engenders threats to traffic safety. Owing to the lack of exposure data, the detection of WWR and its relationship with the built environment (BE) factors remain unclear. Accordingly, this study fills the research gaps by proposing a WWR detection framework based on bike-sharing trajectories collected from Chengdu, China. Moreover, this study adopts Negative Binomial-based Additive Decision Tree to investigate the impacts of built environment on WWR frequencies. Results reveal that (1) WWR distribution is unaffected by different periods in a day; (2) road length is more influential than road level and road direction in WWR occurrence; (3) company, bus stop, subway station, residence, and catering facility are primary contributors affecting WWR behavior during peak hours, whereas education becomes an emerging influential variable during nonpeak hours; and most importantly, (4) these variables clearly present non-linear effects on the WWR frequencies. Therefore, geographically differentiated policies should be adopted for bicycle safety improvement.
Effects of built environment on bicycle wrong Way riding behavior: A data-driven approach
2020-05-24
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch