With the booming of Big Data and the Internet of Things, various urban networks have been built based on intercity flow data, and how to combine them to learn a more comprehensive understanding of mega-city regions is becoming more and more indispensable. In this paper, we designed a graph-based multi-view clustering method based on graph learning to explore the mega-city region structures from multi-source data. An example of clustering analysis consists of the people flow network, cargo flow network, and information flow network, covering 88 cities from Beijing, Tianjin, Hebei Province, Shandong Province, Henan Province, Jiangsu Province, Anhui Province, Shanghai, and Zhejiang Province in China is used to illustrate the applicability of the idea in super mega-city region scale studies. Utilizing the proposed clustering method, a unified network representation is calculated, and 5 mega-city regions, Beijing-Tianjin-Hebei Cluster, Henan Cluster, Shandong Cluster, Shanghai-Jiangsu-Anhui Cluster, and Zhejiang Cluster, are detected based on intercity flow networks. City-to-city flows, including Luan-Taizhou, Lianyungang-Chuzhou, and Xuzhou-Bengbu of the people network, Shanghai-Hangzhou, Suzhou-Shanghai, and Shanghai-Ningbo of the cargo network, Shanghai-Hangzhou, Bozhou-Jinhua, and Huaibei-Bozhou of the information network, are suggested to be further enhanced to facilitate the ongoing nationwide constructions of urban agglomerations in China. The multi-view clustering method proved to be a helpful calculation framework for mega-city region analysis, which would also be considered as a substantial foundation for further urban explorations with more advanced graph learning techniques.
Read full abstract