Understanding shared micromobility travel patterns through a deep embedded clustering approach

Abstract

Shared micromobility services are rapidly expanding across North America, driving the need for improved infrastructure and policy adaptation. Understanding the spatio-temporal travel patterns associated with shared micromobility is essential for effective urban planning and seamless integration into existing transportation networks. However, traditional travel behavior studies, which rely on costly and time-intensive surveys, often fail to capture the granular movement patterns of shared micromobility users. This study employs a Deep Embedded Clustering (DEC) model to analyze shared micromobility trip data, incorporating multiple open-source datasets, including points of interest (POI), transit stops, and weather data. By integrating an autoencoder with k-means clustering, the DEC model is able to effectively uncover distinct travel patterns. A case study in Louisville, Kentucky, reveals significant seasonal and hourly variations in shared micromobility usage. Trip hotspots are concentrated around bars, restaurants, and university campuses, highlighting strong preferences for leisure and education-related travel. Correspondingly, most clusters are associated with Food and Beverage and Transportation POIs, followed by Public Amenities, with one distinct cluster reflecting dedicated education-related travel patterns. These findings provide valuable insights for urban planners and operators, emphasizing the need to enhance shared micromobility regulation and operation and foster better multi-modal connectivity within urban transportation systems.