Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Uncover the nature of overlapping community in cities

Peng Luo, Di Zhu

TL;DR

This paper tackles the problem of overlapping urban communities and their socio-economic implications by introducing a physics-aware graph deep-learning framework. It introduces the Geospatial Graph Affiliation Generation model (GAGM) combined with Graph Convolutional Networks (GCN) to infer node-level community affiliations from large-scale mobility data, treating community membership as a latent affiliation matrix $F$. The study identifies 10 overlapping urban communities in TCMA, linking the overlap index to urban function diversity via POI entropy, to income segregation (weekday nadir, weekend peak), and to racial distributions (White predominance with minority underrepresentation), revealing how overlaps illuminate complex socio-economic dynamics. The findings offer a geospatial perspective for urban planning, suggesting that enhancing shared spaces and mixed-use areas could mitigate segregation and improve cross-community interactions. Overall, the work demonstrates that overlapping community structure is a central, measurable driver of urban socioeconomic patterns, with implications for inclusive design and resource allocation.

Abstract

Urban spaces, though often perceived as discrete communities, are shared by various functional and social groups. Our study introduces a graph-based physics-aware deep learning framework, illuminating the intricate overlapping nature inherent in urban communities. Through analysis of individual mobile phone positioning data at Twin Cities metro area (TCMA) in Minnesota, USA, our findings reveal that 95.7 % of urban functional complexity stems from the overlapping structure of communities during weekdays. Significantly, our research not only quantifies these overlaps but also reveals their compelling correlations with income and racial indicators, unraveling the complex segregation patterns in U.S. cities. As the first to elucidate the overlapping nature of urban communities, this work offers a unique geospatial perspective on looking at urban structures, highlighting the nuanced interplay of socioeconomic dynamics within cities.

Uncover the nature of overlapping community in cities

TL;DR

This paper tackles the problem of overlapping urban communities and their socio-economic implications by introducing a physics-aware graph deep-learning framework. It introduces the Geospatial Graph Affiliation Generation model (GAGM) combined with Graph Convolutional Networks (GCN) to infer node-level community affiliations from large-scale mobility data, treating community membership as a latent affiliation matrix . The study identifies 10 overlapping urban communities in TCMA, linking the overlap index to urban function diversity via POI entropy, to income segregation (weekday nadir, weekend peak), and to racial distributions (White predominance with minority underrepresentation), revealing how overlaps illuminate complex socio-economic dynamics. The findings offer a geospatial perspective for urban planning, suggesting that enhancing shared spaces and mixed-use areas could mitigate segregation and improve cross-community interactions. Overall, the work demonstrates that overlapping community structure is a central, measurable driver of urban socioeconomic patterns, with implications for inclusive design and resource allocation.

Abstract

Urban spaces, though often perceived as discrete communities, are shared by various functional and social groups. Our study introduces a graph-based physics-aware deep learning framework, illuminating the intricate overlapping nature inherent in urban communities. Through analysis of individual mobile phone positioning data at Twin Cities metro area (TCMA) in Minnesota, USA, our findings reveal that 95.7 % of urban functional complexity stems from the overlapping structure of communities during weekdays. Significantly, our research not only quantifies these overlaps but also reveals their compelling correlations with income and racial indicators, unraveling the complex segregation patterns in U.S. cities. As the first to elucidate the overlapping nature of urban communities, this work offers a unique geospatial perspective on looking at urban structures, highlighting the nuanced interplay of socioeconomic dynamics within cities.
Paper Structure (11 sections, 7 equations, 5 figures)

This paper contains 11 sections, 7 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Overlapped network structures in cities
  3. Mapping the overlapping communities
  4. Overlapping socioeconomics in cities
  5. Urban functions explained by overlapping communities
  6. Explaining income segregation with overlapping communities
  7. Minority groups are experiencing limited opportunities
  8. Disjoint vs. Overlapping communities
  9. Method
  10. Datasets
  11. Detect community structure using graph deep learning model

Figures (5)

  • Figure 1: The overlapping nature of communities:
  • Figure 2: Overlapping Communities During Weekends: This figure illustrates the overlapping communities observed during weekends, highlighting three key aspects: (a) the human flows at CBG levels during weekends, (b) the detected ten communities and their associated overlapping patterns; (c) a detailed examination of the overlapping pattern between two specific communities, identified as $Com_a$ and $Com_b$ in (b).
  • Figure 3: Overlapping Index and Urban Functions: (a) The spatial distribution of POI entropy at CBGs level; (b) A scatter plot of the overlap index and POI entropy.The POI entropy is merged on average according to the overlap index; (c) Pearson's correlation between the overlap index and the number of different POIs; (d) Scatter plots illustrating the relationship between POI numbers and the overlap index. Most POI types show a significantly positive relationship (d), including dining, financial, and entertainment, while an exception is found in educational POIs (e), which exhibit a negative correlation with the overlap index.
  • Figure 4: Overlapping index and income. The scatter plot depicts the relationships between the overlap index and household median income.The CBGs level household median income is merged on average according to the overlap index. It reveals a significantly positive correlation during weekends and a negative correlation during weekdays. The map illustrates the spatial distribution of household income, indicating that high-income CBGs are primarily located outside urban centers, where there is a highly diverse human mobility during weekends. Conversely, low-income CBGs tend to be situated in downtown areas, which experience high diverse human mobility during workdays.
  • Figure 5: Overlapping index and race. The figure presents an analysis of the relationship between the overlap index and the percentage of population for different racial groups. (a) reveals a significant positive correlation for the white population, while a negative correlation is observed for the black population; (b) shows the Pearson correlation coefficients between the overlap index and the population percentage of four racial groups are displayed, showing that only the white population exhibits a positive correlation with the overlap index; (c) and (e) map the spatial distribution of the overlap index for the white group, while (d) and (f) represent the black group.