Shared micromobility services provide a flexible and sustainable alternative for urban and suburban mobility, especially for the first and last mile problems. It is crucial to understand the mobility patterns in the operating areas to provide the best experience for users of shared micromobility services. Effective demand modeling can provide data to test and evaluate fleet configurations in different but relevant mobility scenarios. This study utilizes statistical and deep learning techniques for demand modeling of a shared e-moped service operating in the urban and suburban areas of Stuttgart, Germany. It addresses the research gap of demand modeling in both urban and suburban areas and compares the used machine learning approaches. The results show that deep learning models, specifically LSTMs, provide the best results, outperforming the baseline and traditional models on most metrics. Using grid and Voronoi systems and community clustering to aggregate trips and reduce spatial complexity led to better results for all evaluated models. Findings contribute to the demand modeling of micromobility services operating in urban and suburban areas with sparse data.
If this work is useful or relevant to your research, please acknowledge our contributions by citing our work:
@article{roehner2025demandgeneration,
title={Data-Driven Approaches to Micromobility Demand Modeling},
author={Ruben Röhner, Damir Ravlija, Ingo Trautwein and Mirko Sonntag},
journal={},
year={2025}
}