Exploring e-scooters as first- and last-mile with Gaussian mixture and machine-learning models

Abstract

E-scooters are emerging as a key urban transportation option that can improve public transportation accessibility and address first- and last-mile issues. However, their utility is uncertain when destinations do not align with transit routes. This study examines whether e- scooters complement or substitute public transit by analyzing travel segments from trip origins to transit stations and from stations to final destinations. Using GPS-based mobility data, Gaussian mixture model, and explainable machine learning, we identified seven distinct patterns of e-scooter use based on population and land-use characteristics. Most patterns show e-scooters complement transit by connecting underserved residential, commercial, and green areas to transit hubs. The results indicate substitution-like patterns in low-density suburban areas characterized by low bus service frequency and limited metro accessibility. These findings highlight the role of e-scooters as a complementary transport mode when conventional public transport coverage is weak, suggesting their relevance for first- and last-mile integration within mobility-as-a-service.