From bicycle sharing system movements to users: a typology of Vélo’v cyclists in Lyon based on large-scale behavioural dataset

Marie Vogel, Ronan Hamon, Guillaume Lozenguez, Luc Merchez, Patrice Abry, Julien Barnier, Pierre Borgnat, Patrick Flandrin, Isabelle Mallon, Céline RobardetJournal of Transport Geography 2014 · DOI

TL;DR

This paper leverages a large-scale dataset of bicycle sharing system (BSS) movements in Lyon to develop a novel user typology based on actual cycling behavior, moving beyond traditional demographic or survey-based classifications. This approach offers a more nuanced understanding of BSS users and their integration into urban mobility, providing valuable insights for urban planning and pro-cycling strategies.

What Problem Does This Paper Address?

While bicycle sharing systems (BSS) have rapidly proliferated globally, research has often focused on system-level analysis or relied on self-reported data from surveys. This leaves a gap in understanding the actual, individual-level mobility patterns of BSS users and how these patterns relate to user characteristics. Existing typologies of cyclists often rely on motivations or general practice, but rarely on detailed, real-world movement data. This paper aims to bridge this gap by using a rich dataset of BSS movements and user data to create a data-driven typology of BSS users, specifically focusing on the Vélo'v system in Lyon. This allows for a deeper understanding of user behavior, the integration of BSS into multimodal travel, and the conditions that foster wider bicycle use in cities.

Methods

The core of this research lies in a data-driven approach to user segmentation. The authors employed a K-means clustering algorithm to group Vélo'v users based on a vector of 21 attributes quantifying their intensity and regularity of use. These attributes capture weekly activity (e.g., average trips per weekday, weekend use) and annual activity (e.g., trips per month). To ensure robust interpretation, the authors used statistical criteria like the Average Means Silhouette (AMS) and explored different numbers of clusters (3, 7, 9, and 12) before settling on nine distinct user clusters. They opted to use the original normalized attributes directly for clustering rather than PCA axes to facilitate straightforward interpretation of the resulting classes. The analysis also involved comparing the profiles of these annual user clusters with short-term (daily and weekly) users, and examining spatial distribution patterns.

Datasets

The primary dataset used in this study is a comprehensive collection of bicycle movement and anonymous user data for the Vélo'v system in Lyon, France, for the calendar year 2011. This dataset was provided by the operator and includes:

  • Bicycle movements: Details on the start and end time and location of each journey, indexed to a specific membership type (annual, weekly, or daily).
  • User data (for annual members): Anonymous information including age, gender, and postcode.

The authors excluded journeys shorter than 3 minutes and longer than 24 hours, as well as "suspicious" users (24 in total) with unrealistic practices. The dataset comprises approximately 6.5 million movements. This dataset is notable for its scale and the integration of both movement and user data, allowing for individual-level analysis.

Key Insights and Ideas

  • Behavioral Typology is Key: The most significant insight is that a user typology based on actual movement patterns (intensity and regularity) provides a richer understanding than demographic data alone. This moves beyond "who" users are to "how" they use the system.
  • BSS as Integrated Mobility: The study highlights that BSS users are not a monolithic group but rather exhibit diverse integration with other transport modes, ranging from dedicated users to those using it as a flexible supplement.
  • Gender and Intensity Correlation: A clear trend emerges where higher intensity of BSS use is correlated with a higher proportion of male users, though this is nuanced by specific user categories.
  • Spatial Distribution is Less Differentiating: Contrary to some expectations, the study found that users' residential location (postcode) was not a strong predictor for distinguishing between user clusters, suggesting BSS use is more about mobility patterns than strict residential proximity to stations.
  • Short-term vs. Annual User Differences: A stark contrast exists between annual members, who tend to use BSS for commuting during weekdays, and daily pass holders, who are more likely to use it for leisure activities, particularly on weekends and at night.

Key Results

  • Nine User Typologies: The clustering analysis identified nine distinct user profiles, which were further grouped into four broader categories: "Users of heart" (highly intensive and regular), "Assiduous users" (intensive and regular, sometimes seasonal), "Multimodal users" (low but regular use, balanced gender ratio), and "Sporadic users" (low intensity, often weekend or occasional use).
  • Usage Intensity Varies Dramatically: The number of trips per year per user ranged from as few as 2 to over 690, with the top four classes (29% of users) accounting for nearly 75% of all journeys.
  • Gender Skew in Intensive Use: "Users of heart" and "Assiduous users" were significantly male-dominated (around 73% and 60% male, respectively). The "Multimodal users" category showed a much more balanced gender ratio (0.82 W/M).
  • Age Profiles Differ by Intensity: While overall age differences weren't stark, "Assiduous users" in class E (intensive but part-time) were notably younger, likely students.
  • Short-term Users for Leisure: Daily pass holders primarily used the system for leisure, with journeys concentrated in the latter half of the day and at night, and significantly longer average trip durations compared to annual members.

Limitations and Future Work

The authors acknowledge two main limitations:

  • Data Granularity for Short-Term Users: The dataset provided limited demographic information for daily and weekly members, making it difficult to fully characterize their profiles beyond their movement patterns.
  • Focus on Behavior, Not Incentives: The study's focus on actual cycling patterns means it doesn't directly capture the motivations, incentives, or preferences that drive users' choices.

Future work is planned to address these limitations. This includes:

  • Surveys and Interviews: Conducting targeted surveys and interviews with both BSS users and non-users to gather qualitative data on motivations, perceptions (e.g., safety), and multimodal travel choices.
  • Broader Program: This study is presented as the first phase of a larger research program aiming to explore BSS use more thoroughly, incorporating spatial distribution of movements and frequency analysis at different time scales.
  • Cross-City Comparisons: The authors intend to compare the Vélo'v system in Lyon with BSS in Valencia (Spain) and Montreal (Canada) to identify commonalities and differences in user behavior and usage patterns, potentially highlighting the influence of local socio-historical contexts.

Summary updated

BibTeX
@article{vogel2014from,
  author = {Marie Vogel and Ronan Hamon and Guillaume Lozenguez and Luc Merchez and Patrice Abry and Julien Barnier and Pierre Borgnat and Patrick Flandrin and Isabelle Mallon and Céline Robardet},
  title = {{From bicycle sharing system movements to users: a typology of Vélo’v cyclists in Lyon based on large-scale behavioural dataset}},
  journal = {Journal of Transport Geography},
  year = {2014},
  doi = {10.1016/j.jtrangeo.2014.07.005},
}
Figure