Abstract
Data from GPS-based fitness tracker apps have become a prominent source for studying cycling behavior. This type of crowd-sourced data provides larger datasets than were previously attainable by travel surveys and cyclist counts, which allows for the comparison of trip characteristics between geographic regions and the study of temporal trends in bicycle ridership. Researchers acknowledge that different types of biases come with tracking data from fitness tracker apps, such as self-selection bias, which is partially based on different target audiences among fitness tracker apps. This begs the question as to whether the behavior of cyclists, and thus the characteristics of routes traveled, varies among the apps. To provide a first insight into this question, this research analyzes trips reported on three fitness tracker apps, Bikemap, Endomondo, and MapMyFitness, for South Florida (Miami-Dade, Broward, and Palm Beach counties) and North Holland. Comparison of trip characteristics is made among the three apps and across both study regions. Results show that cycling behavior observed in the three apps is similar relative to a set of control trips in each region (e.g. fewer primary roads than reference trips observed), but that there are some pronounced differences in trips recorded with the different apps between both regions. This suggests that geographic region plays a role in how trip characteristics recorded on different apps compare to each other, demonstrating the presence of an additional aspect of geographic bias in crowd-sourced cycling data.
Acknowledgements
The authors would like to acknowledge Levente Juhász for providing the Flickr image extraction script and invaluable help with programming related questions. Angela Schirck-Matthews has been awarded an international research scholarship for this project by the Austrian Marshall Plan Foundation, which is gratefully acknowledged.
Disclosure statement
No potential conflict of interest is reported by the authors.
Data availability
The CSV files with attributes of routes that were analyzed in this study as well as Python scripts for extracting trip points from MapMyFitness and Bikemap can be found on GitHub at https://github.com/Mathwiz961/AppCompare.
Notes
6 https://floridadep.gov/