Understanding bike share cyclist route choice using GPS data: Comparing dominant routes and shortest paths

This paper investigates cyclist route choices using global positioning system (GPS) data collected from 750 bicycles in Hamilton, Ontario's bike share system - SoBi (Social Bicycles) Hamilton. A dataset containing 161,426 GPS trajectories describing observed routes of cyclists using SoBi bikes over a 12-month period (April 1, 2015 to March 31, 2016) is used for analysis. This study groups trips by origin-destination hub pairs and uses a GIS (geographic information system)-based map-matching algorithm to generate routes along with attributes such as length, number of intersections, number of turns, and unique road segments. Unique routes and their use frequencies are extracted from all the hub-to-hub trips using a GIS-based link signature extraction tool developed for this research. The most popular routes between hubs taken by cyclists are then identified as dominant routes and their attributes are compared to those of corresponding shortest path routes derived by minimizing distance traveled. The comparison finds significant differences in multiple attributes, and demonstrates that dominant routes are significantly longer than their shortest distance counterparts, suggesting that cyclists are willing to detour for routes characterized by positive features such as bicycle facilities and low traffic volumes. Detouring does, however, come at a cost - increases in number of turns and number of intersections. This research not only enhances our understanding of cyclist route preferences within a bike share system, it also presents a GIS-based approach for identifying potential locations for future bike facilities based on such preferences.