- This event has passed.
Thesis Defence: Modelling Cyclists Trip Purposes and Route Choices: Application of Machine Learning Models Using GPS Data
November 23 at 12:00 pm - 3:00 pm
Maryamossadat Zakeri, supervised by Dr. Mahmudur Fatmi, will defend their thesis titled “Modelling Cyclists Trip Purposes and Route Choices: Application of Machine Learning Models Using GPS Data” in partial fulfillment of the requirements for the degree of Master of Applied Science in Civil Engineering.
An abstract for Maryamossadat Zakeri’s or thesis is included below.
Defences are open to all members of the campus community as well as the general public. Please email email@example.com to receive the Zoom link for this defence.
This thesis aims to uncover bicyclists’ behaviours, specifically their route choices and trip purposes. The study employs data from GPS records of a dock-less bikeshare service (DBS) in Kelowna, Canada. A significant contribution of this research is the application of machine learning (ML) algorithms to infer trip purposes from GPS records, in the absence of direct survey data. Notably, the findings reveal that mandatory trips to work and educational locations dominated DBS usage. Additionally, spatio-temporal distribution of the trips for distinct purposes is provided. For instance, the analysis clarified that trip objectives such as shopping and dining were densely clustered around the urban core area, while return-home trips displayed a more dispersed pattern.
Another primary focus of this research is the design of a data-driven choice set generation method for cyclists’ route choice behaviour modelling. This approach utilizes machine learning algorithms to create feasible choice sets from trip records, followed by a mixed logit model to distinguish the specific preferences and patterns in cyclists’ behaviours. A notable finding from the model validation reveals the superiority of the GMM clustering method (developed in this study) over k-means and random choice set generation methods.
The thesis also identifies the impact of various factors such as the built environment, land use, accessibility, and temporal attributes on bicycling behaviour. Among the insights, the models underline the cyclists’ tendency for routes with specialized cycling infrastructure, such as cycle tracks and lanes, emphasizing the need for increased infrastructure investment. Furthermore, there exists behavioural heterogeneity, as exemplified by the mixed reactions of cyclists to routes near transit stops, suggesting both opportunities for mode transfer and potential disadvantages due to interactions with larger vehicles. Therefore, this research as well highlights that future planning and policy formulation should incorporate these diversities to optimize the use of bikes and shared services.