Categorizing three active cyclist typologies by exploring patterns on a multitude of GPS crowdsourced data attributes

This paper tackles the problem of characterizing cyclist typologies using a data set of GPS traces. The data is crowdsourced and consists of 29,431 traces recorded in the city of Bologna, Italy, during the rush hours from 7 am to 10 am on weekdays and during the months of April through September in 2017. The different criteria used to group the heterogenous behavior of cyclists into separate categories have already been described in literature and those studies were generally based on the stated preference in interviews or upon observing a small sample of the population. The novelty of this study lies in the clustering of a data set of 2135 cyclists that included GPS traces, which is equivalent to a revealed survey. Furthermore, refined pre-processing of the GPS traces allows the determination of dynamic attributes, a comparison of the chosen path with respect to the shortest path and the evaluation of other specific trip attributes, which are either difficult or impossible to assess by a classical interview. The applied clustering process leads to three main cyclist typologies, such that each type is characterized by different trip attributes and behaviors involving safety, riskiness, precaution, inexperience, knowledge, fear and hastiness: risky and hasty, sly and informed and inexperienced and inefficient.