Detecting The Overlapping Community Structure In Urban Road Networks By Using Hierarchy Information

The research of road network structure, which are the main carrier of city transportation, can improve the decision-making quality of urban planning and transportation management. The automatic generalization of road networks has been regarded as an important part of automatic map generalization. Considering the community structure is the basic foundation of road networks automatic generalization, researches on the community structure partitioning are extremely important for the structural analysis of the road networks and the automatic map generalization.Previous studies on road network structure are mostly based on the theory of complex network. However, it is not appropriate for using the original method of complex network directly to do the research of road network, because the road network has its own inherent characteristics. Considering hierarchy is the most obvious characteristic of road network, this paper aims to partition road network based on the complex network theory and road hierarchy, which can be obtained using some verified methods.This paper proposes an improved method, which detects non-overlapping communities from road networks using local structure optimization and ant-based random walk method. To avoid biased conclusions, a classic network and some real road networks are selected as experimental data. By comparing the result of partition with other algorithms, the experiment shows this improved algorithm is both effective and reasonable for discovering communities.This paper proposes a practical overlapping community structure partitioning model for road networks by studying the inter-relationships between the hierarchies and the overlapping structures of road networks, together with the idea of overlapping community structure partitioning method on the condition of overlapping nodes. It takes dual graph of stroke as the research complex network structure, and then started with the partitioning of the non-overlapping community structures as the initial community structure, then marked strokes with overlapping boundary nodes for each pair of non-overlapping communities to get the overlapping communities of road network.The algorithms included in this paper are realized with C++ language. With the comparison between the experiment results of this method in this paper and other algorithms, the results show that the classification of road networks by hierarchy gets better results in overlap modularity. Also the visualization of these results accords with natural phenomena in reality, which proves the rationality and the good feasibility of this method.