Road networks: A new filtering approach to extract backbones using community structure

Road networks represented as a networked system of edges and nodes have raised considerable interest in the literature of networks, leading to numerous studies throughout the years. In these networks, the junctions where roads intersect are represented by nodes while the road segments connecting them represent the edges. Road networks in major cities and metropolises all over the world are very large-scale dense networks. Thus, the need to identify the most important roads that influence both vehicle and pedestrian flows, urban crime, collective dynamic behavior, land-use separation and retail vitality. In this context, in this paper, we propose a new method to extract the skeleton or the backbone of cities. The so-called "Link Motif-Betweenness skeleton (LMB)" aims to select the links belonging to the larger number of motifs in the network and also having high betweenness. The LMB selects links separately from two sub-networks: from the local sub-network which is formed only from the intra-community links to select edges with a local influence in the various communities, and also from the global sub-network which is formed from the inter-community links of the network to select edges with a global influence all over the network. The proposed method is tested on road networks in the five largest cities of France: Paris, Marseille, Lyon, Toulouse and Nice. Experimental results show the efficiency of the LMB in terms of preserving relevant roads as compared to some alternative methods.