Evaluating bicycle lanes and stations safety in Barcelona: a Bayesian network approach to analyse injury severity

Abstract

Bicycle infrastructures are increasingly expanding in various regions with the aim of providing sustainable transportation options. In Barcelona, Spain, the rise in bicycle usage is accompanied by the development of appropriate supporting infrastructures. This study utilises Geographic Information System (GIS) to map out the bicycle lanes, stations, and zones. Furthermore, data on bicycle crashes in Barcelona in 2019, categorised according to three types of injuries, are combined with information on location, gender, age, and time of day. A Tree Augmented Naïve tree (TAN), which falls under the Bayesian network family, is then employed to classify the potential factors influencing these incidents. The findings reveal that individuals aged 26–50 years are the most vulnerable to experiencing slight or severe injuries while cycling, particularly during morning and evening hours. Additionally, bicycle lanes and stations located in zones with a restricted speed limit of 30 km/h exhibited fewer instances of crash injuries compared to other types. The TAN approach demonstrated robust predictive performance when applied to similar datasets.