Spatial structure of Spatial Interaction: Using Graph structural information in Modelling Bipartite Networks

Keywords: spatial interaction, spatial structure, urban flow, origin-destination, comparing spatial patterns, network structure
Session Type: Virtual Paper Abstract
Day: Friday
Session Start / End Time: 2/25/2022 12:40 PM (Eastern Time (US & Canada)) - 2/25/2022 02:00 PM (Eastern Time (US & Canada))
Room: Virtual 18

Abstract

Spatial Interaction Models (SIM) have been widely used to model migration, urban commuting, and trade flows. However, SIMs are used to model processes with a very typical structure. It is unclear as to whether current methods can adequately account for non-traditional types of network, such as bipartite ones. Furthermore, the models are usually validated by standard predictive measures, that aims to evaluate the predictive performance for each flow in isolation, which does not allow us to evaluate how well the models capture the pattern of flows. Finally, we face a vague explanation of what spatial structure is, how it is conceptualized and incorporated into models. In this work, we explore the concept of spatial structure and draw an inference about its representation in the current modeling framework. We then explore the potential of graph structure measure, specifically PageRank, to provide a general measure of spatial structure, and examine its performance as an alternative measure of accessibility in SIMs. We do this for two different types of networks: unipartite and bipartite, and compare models with standard predictive performance methods as well as comparing their spatial pattern reconstruction. We find that PageRank accounts for changes at both the local and global scale. It may provide a more useful general measure of spatial structure for typologically different SIMs and can yield estimates that are superior to traditional measures of accessibility. Overall, this work encourages us to think more critically about spatial structure in SIMs and widen our ideas of what constitutes ”good performance”.