Bicyclists’ travel behavior analysis: modeling bicycling demand, speed choice, destination choice, and usage of shared mobility services

This thesis develops advanced modeling techniques to investigate bicycling demand and bicyclists’ user behavior including their speed choice, destination choice, and reasons to choose dockless bikeshare service (DBS). This study utilizes data from automatic bicycle counters, GPS records, and DBS user survey from the City of Kelowna. One of the major contributions of this thesis is to adopt a machine learning (ML) algorithm to improve the trip identification procedure using the GPS data. Another key feature of this thesis is to develop latent segmentation-based (LS) models to capture heterogeneity. For example, the destination choice analysis adopts a random parameter LS logit model to address multidimensional heterogeneity by assuming a discrete (i.e. inter-segment heterogeneity), and continuous (i.e. intra-segment heterogeneity) distribution of the parameters. This thesis examines the effects of weather, built environment, traffic, land use, neighborhood, socio-demographic, accessibility, and trip attributes on bicycle usage. In case of the model results, the proposed advanced models show improved goodness-of-fit measures. For instance, speed choice model validation results suggest that ML and LS modeling approach outperforms other methods in-terms of Welch’s t-test and Mean Absolute Deviation respectively. The model results provide important behavioral insights which are expected to assist in developing effective policies to promote bicycling more. For example, the model results suggest that bike index (BI) is an important target variable to increase bicycling demand and speed. Individuals are more likely to choose DBS as cheapest option while residing in bicycle-friendly neighborhoods with good transit accessibility. The model results also confirm the existence of multidimensional heterogeneity. For instance, trip destinations closer to designated dockless bike return areas (i.e. havens) show inter-segment heterogeneity. Specifically, urban trips are more likely to be destined closer to the havens; whereas, suburban trips are less likely to be destined closer to the havens. Intra-segment heterogeneity is also confirmed while this variable is tested for BI – i.e. a higher BI indicating bicycle-friendly environment increases the probability for suburban trips to be destined closer to the havens. Finally, heterogeneity captured in this research needs to be accommodated within the policies to ensure effective usage of bicycle and newer shared services.Applied Science, Faculty ofEngineering, School of (Okanagan)Graduat