Detecting urban structure through inter-city human interactions in China

We combine novel inter-city human interaction data with traditional node attribute data to explore how the Chinese urban network is structured, and how this is associated with the economic performance of cities in an increasingly urbanising China. We then employ well-established unsupervised machine learning algorithms to cluster the Chinese cities based on these network variables and create urban typologies based on mid-term migration patterns. We identify 7 clusters of cities with shared network characteristics and contrast their economic performances based on more traditional data. The most disconnected, peripheral cities are also shrinking with a negative population difference, but are not necessarily the weakest economically, with a reliance on primary industry