City Motifs as Revealed by Similarity Between Hierarchical Features
Several natural and theoretical networks can be broken down into smaller portions, or subgraphs corresponding to neighborhoods. The more frequent of these neighborhoods can then be understood as motifs of the network, being therefore important for better characterizing and understanding of the overall structure. Several developments in network science have relied on this interesting concept, with ample applications in areas including systems biology, computational neuroscience, economy and ecology. The present work aims at reporting an unsupervised methodology capable of identifying motifs respective to streets networks, the latter corresponding to graphs obtained from city plans by considering street junctions and terminations as nodes while the links are defined by the streets. Remarkable results are described, including the identification of nine stable and informative motifs, which have been allowed by three critically important factors: (i) adoption of five hierarchical measurements to locally characterize the neighborhoods of nodes in the streets networks; (ii) adoption of an effective coincidence methodology for translating datasets into networks; and (iii) definition of the motifs in statistical terms by using community finding methodology. The nine identified motifs are characterized and discussed from several perspective, including their mutual similarity, visualization, histograms of measurements, and geographical adjacency in the original cities. Also presented is the analysis of the effect of the adopted features on the obtained networks as well as a simple supervised learning method capable of assigning reference motifs to cities.