Exploring dynamic urban mobility patterns from traffic flow data using community detection

Abstract

The rise of data from smart city services and the emergence of advanced algorithms have emphasized the need for a deeper understanding of the underlying patterns of urban mobility and the potential opportunities for more efficient urban planning and policymaking. By identifying communities with similar mobility patterns, researchers can gain better insights into how people move within and between different regions. Traditional community detection methodologies mainly focused on identifying geographic communities defined by shared locations. However, this perspective overlooks the broader definition of communities. Communities can also emerge from shared social interests, collective actions, and activity patterns. This implies that geographically disparate areas might exhibit similar patterns, which is particularly relevant in the study of mobility pattern similarity. Rather than focusing on regions with strong spatial interactions, this study aims to identify regions that show more similarity to each other than to other areas. Such similarities may indicate parallel urban functionalities, which are essential for effective urban planning and policymaking. To bridge this gap, our study introduces a customized community detection algorithm that employs Dynamic Time Warping (DTW) to quantitatively assess the similarity in mobility patterns between different communities. This advanced approach not only improves the identification of comparable mobility patterns but also demonstrates remarkable flexibility, broadening its application to various other social phenomena. The results demonstrate the effectiveness of the proposed model in capturing complex mobility patterns across different locations and days of the week.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 15 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.