Modeling Destination Choice Behavior of the Dockless Bike Sharing Service Users

Abstract

This study investigates trip-level destination choice behavior of users of the dockless bike sharing service (DBS). A random parameter latent segmentation-based logit (RPLSL) model is developed utilizing the DBS users’ trip itinerary data for Kelowna, Canada. The RPLSL model captures multi-dimensional heterogeneity such as inter-segment and intra-segment heterogeneity. The model is developed at a micro-spatial resolution which is defined as the bicycle analysis zone. One of the key features of this study is to test the interdependencies between the origin and destination of a trip using their built environment attributes. Model results suggest that segment 1 is more likely to include trips originating from the urban neighborhoods; whereas, segment 2 includes trips originating from the suburban neighborhoods. The parameter estimation results reveal that DBS trips are more likely to be destined to locations with longer length of cycle tracks, higher employment density, and that are closer to the Central Business District and bus stops (i.e., within 500 m). The model confirms multi-layer heterogeneity. For instance, trips originating from the urban areas in segment 1 are more likely to be destined to destinations within 500 m of the designated bike return sites (i.e., havens). In contrast, shorter trips originating in suburban areas in segment 2 show a negative relationship. Interestingly, a bike-friendly environment might increase the attractiveness of destinations closer to havens, even for the trips originating in suburban areas. The findings of this study will assist in developing policies and infrastructure investment decision making at the destination locations to promote DBS usage.

Joint Econometric Model Framework for Transportation Network Company Users’ Trip Fare and Destination Choice Analysis

“…The result indicates income inequity in the adoption of TNC for mobility needs in Chicago and other urban regions, as found by other studies (23,65,66). As expected, on weekdays, a census tract with higher employment density is likely to attract more TNC trips (31,65).…”

“…Destination selection behavior has been examined in multiple ride-sharing domains including bicycle-sharing system, taxi, TNC, and shared autonomous vehicle (SAV) (30)(31)(32)(33)(34)(35)(36)(37). The preferred approach employed at the disaggregate level is the MNL model based on the random utility maximization approach (30).…”

Developing vehicular and non-vehicular trip generation models for mid-rise residential buildings in Kelowna, British Columbia: Assessing the impact of built environment, land use, and neighborhood characteristics

“…These parameters are estimated using the maximum likelihood method. The goodness-of-fit measure of the model is evaluated using a log-likelihood function, Akaike Information Criteria (AIC), Bayesian Information Criteria (BIC), and adjusted pseudo rho-squared value (Orvin & Fatmi, 2020a).…”

A Systematic Literature Review on Machine Learning in Shared Mobility

“…Linear Model [185,186,187] Tree-based Ensembles [184,186,188,189] Neural Networks [190,191,192] Scooter Dockless OD + Neural Networks [193] For example, demand in one area can be affected by traffic in other areas, and external factors such as weather, events, and holidays can have an impact on demand throughout all regions. Despite significant research in traffic forecasting, spatio-temporal forecasting remains an area of ongoing study.…”