Exploring micromobility choice behavior across different mode users using machine learning methods

Abstract

In an effort to capture travelers’ propensity towards micro-mobility options, a consumer survey was designed and conducted in the state of Florida in Fall 2021. In addition to collecting socioeconomic, demographic, attitudinal, and trip-related information, stated-preference scenarios were presented to the respondents, in which they were asked to choose between their current mode, and three different micro-mobility alternatives, namely: e-scooter, e-scooter + public transit, and moped. A machine learning classification model, the tree-based Extreme Gradient Boosting algorithm was applied to study users’ mode choice toward micromobility options given its non-parametric nature and high predictive power. SHAP values were then used to analyze the contributing factors for each of the micro-mobility options. In addition, Local Interpretable Model-agnostic Explanations (LIME) was employed to interpret and validate the SHAP findings at the individual prediction level. Model results show that age, car-oriented attitudes, lack of familiarity/previous experience, and lack of appropriate infrastructures were the major barriers to choose micro-mobility services. Such services can be suitable alternatives for young people who come from large families or ride-share users who have prior experience with micromobility services. Among different micro-mobility alternatives, mopeds were favored by males and green travelers. It seems that e-scooter + public transit was considered a safe and comfortable option, especially for students and low-income individuals, but generally not favored by travel time-sensitive or green travelers. Finally, e-scooters seem to be a favorable option for younger individuals with short travel distances. Our findings provide additional insights on policies that may help encourage the use of micromobility devices and promote sustainable, affordable, and equitable mobility services.

Keywords

• Stated-preference
• Micro-mobility
• E-scooters
• Attitudes
• Machine learning