Modelling bicycle route choice using data from a GPS-assisted household survey

Ghanayim, Muhammad; Bekhor, Shlomo · 2018 · Crossref

DOI: 10.18757/ejtir.2018.18.2.3228

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates bicycle route choice behavior for commuter trips in the Tel Aviv metropolitan area, addressing the challenge of generating representative route choice sets for modeling. While previous research often relied on non-representative samples or simplified planning networks, this paper utilizes data from a GPS-assisted household travel survey to capture the preferences of a general population. The research aims to estimate route choice models that account for detailed network characteristics and land use variables, while testing the sensitivity of parameter estimates to different choice set generation methods and sizes. The methodology employs a dataset of 545 bicycle trips recorded via GPS devices from 221 individuals during a 2013–2014 survey. These trips were map-matched to a detailed urban network comprising over 127,000 links. To generate alternative routes for each origin-destination pair, the authors applied three shortest-path-based techniques: link elimination, link penalty, and simulation. A final choice set of 20 routes per observation was constructed by combining outputs from these methods and removing routes with more than 80% overlap. The authors also proposed a generalized overlap measure that accounts for specific attributes like bikeway usage, rather than just physical length. Model estimation was performed using Mixed Logit specifications, including Multinomial Logit (MNL), C-Logit, and Path Size Logit (PSL), with utility functions incorporating variables such as total length, length on dedicated bike paths (Category A), length on busy arterials (Category C), average street length, dwelling density, and proximity to sea or parks. The results indicate that cyclists generally prefer shorter routes but are willing to extend their travel distance to utilize separated bike lanes. Specifically, the ratio of actual route length to the shortest route was 1.13 overall, rising to 1.18 in areas with dedicated bicycle facilities. Model estimation revealed that riders significantly avoid busy arterial streets and highways while preferring local streets with fewer intersections and higher residential density. Proximity to pleasant environments, such as the sea or parks, also positively influenced route choice. Among the tested models, the Mixed PSL outperformed Mixed MNL and Mixed C-Logit, achieving a hit-ratio of 69% and a rho-bar squared of 0.316. Sensitivity tests demonstrated that parameter estimates remained robust even when the choice set size was reduced by half, although the standard deviation of coefficients was more sensitive to choice set reduction than the mean estimates. The significance of this work lies in its use of a representative, GPS-based household survey to model bicycle route choice, providing more accurate insights than studies relying on specific population segments. The findings confirm that cyclists prioritize safety and comfort, evidenced by their preference for dedicated lanes and avoidance of high-traffic roads, even at the cost of increased travel distance. Furthermore, the study validates the robustness of route choice models against variations in choice set composition, suggesting that smaller, well-constructed choice sets can yield reliable parameter estimates. This contributes to more effective bicycle network planning by highlighting the importance of infrastructure quality and environmental context in shaping commuter behavior.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• cyclist safety