RN-Cluster: A Novel Density-Based Clustering Approach for Road Network Partition

Abstract

AbstractAs an indispensable part of urban road management, road network partition is essential for the design and planning of road systems. In this paper, we focus on the problem of urban road network partitioning and propose a density-based clustering method RN-Cluster. Specifically, by taking road junctions as vertices, the relationships between road junctions and road sections as edges to construct the road knowledge graph, our method can effectively infer the implicit information of each road or junction. In addition, we propose a measurement of attenuation of neighborhood connectivity (ANC) as the density measurement to select core vertices. Our experiments show that the proposed method can efficiently divide the urban road network into different reasonable areas while achieving state-of-the-art performance on multiple metrics.KeywordsClustering methodSpatial partitionRoad graph