Determinants of pre-ride walking behaviour in dockless shared micromobility
Research in Transportation Business and Management, vol.69, 2026 (SSCI, Scopus)
- Publication Type: Article / Article
- Volume: 69
- Publication Date: 2026
- Doi Number: 10.1016/j.rtbm.2026.101881
- Journal Name: Research in Transportation Business and Management
- Journal Indexes: Social Sciences Citation Index (SSCI), Scopus
- Keywords: Access duration, Decision support system, Shared free-floating, Urban mobility, Walking patterns
- Akdeniz University Affiliated: Yes
Abstract
Walking duration inducing walking behaviour is a key determinant of understanding constraints on the usage of dockless micromobility systems. Thus, to better understand the characteristics of the walking duration of the users of these systems based on the environmental, spatial, and operational determinants, an augmented 17-month dataset with approximately 59 million location telemetry logs and 1 million trip records from 10 distinct cities in Türkiye is used, which is provided by the operator called GEZ. First, a precise matching algorithm, showing users' realized walking durations, is reconstructed, suggesting that 31.5% of trips are initiated with a walk of one minute or less, underscoring a critical behaviour for immediate accessibility. Then, the constructed General Linear Mixed Model demonstrates that walking duration is accounted for by environmental, spatial, and operational factors. More specifically, the findings suggest that longer walking duration is linked to no precipitation and cooler weather conditions under environmental conditions, on the other hand, walking upward always decreases the walking duration on average. Even though the mean walking duration tends to decrease as the number of providers increases, the GLLM model suggests that the walking duration decreases significantly is not guaranteed. Furthermore, the constructed ANN regression model, using environmental, temporal and operational factors, estimates the probability distribution of four distinct walking duration groups with 0.05 MAE, suggesting that under given specific conditions, the operator can predict how each walking duration group is realized.