Using crowdsourced data to map bicycling behaviour. GeoComputation 2019

Growing interest in data for bicycling planning has many jurisdictions aiming to develop a sampling strategy for measuring bicycle counts. However, it can be difficult to determine where bicycling counters should be placed to generate a representative sample of bicycling, and bicycle counts tend to only occur on heavily used bicycle routes. Our goal, is to utilize crowdsourced data on bike ridership, from Strava Metro, and continuous signal processing data mining techniques to map regions of bicycling behaviour in San Francisco – CA, US. Bicycling behaviour regions can be used to stratify bicycle count sampling. Our results indicate it is possible to differentiate bicycling behaviour from Strava and we mapped routes categorized into six unique ridership behaviours including commute to work, commute to school, leisure, 8 am peak , 11 am peak and least utilised. We recommend sampling ridership from each of the categories when developing count programs.