An analysis of urban land subsidence susceptibility based on complex network

The damage wrought by urban land subsidence increases as cities grow, engineering projects expand, and there is more human activity every year. This paper uses the complex networks method with graph neural network learning and regression models to investigate the spatial characteristics and influencing factors of land subsidence during the second phase of the construction of Shenzhen Metro Line 5 in 2019. The outcomes of the experiment reveal that (1) community testing of the settlement network produced nine settlement funnel ranges, and the largest central depth was \(-\) $-$ 24.52 mm/a; and (2) the graph attention network neural network divides the sensitivity of network nodes into three categories based on 15 different influencing factors. Comparison of the land subsidence funnels results reveals a considerable association between the occurrence of settlement during subway construction and both land vegetation and geological types. The study’s findings offer some scientific support for preventive and control measures for land subsidence management in urban design.