Predicting Micromobility Demand in University Campus Environments

Shared Micromobility services such as e-scooters and e-bikes are promising transportation alternatives, but their successful implementation is highly dependent on a comprehensive understanding of the demand patterns of each. Thus, this study applied a hybrid conditional modeling approach that combined a classification and regression model to explore the demand for micromobility and predict ride volumes between zones on a university campus. The results showed that the predictive accuracy of the conditional model was superior to that of the traditional direct prediction model, and an analysis of the feature importance of the traditional model provided deeper insights by identifying the most influential factors in determining ride counts as day of the week, pickup zone, and drop-off zone; holidays and exam schedules had a comparatively minor effect. These results underscore the significance of weekly temporal trends and campus-specific locations in forecasting micromobility demand and offer actionable insights for optimizing campus transportation planning and operations.