Urban construction is believed to exacerbate geological hazards, yet direct evidence remains limited. By examining the changes in disaster susceptibility of urban road collapses in Binjiang District, Hangzhou, for the years 2008, 2016, 2023, and the projected year 2030, we aim to illustrate the impact of engineering construction on geological hazards. Thus, we selected six static factors along with three dynamic factors most reflective of human engineering activities in Hangzhou. To predict disaster susceptibility in 2030, first, we employed a Long Short-Term Memory (LSTM) model to forecast the dynamic factors; Second, by integrating LSTM with Generative Adversarial Networks (GANs) models, we achieved predictions of collapse for 2030. Given the small and scattered nature of collapse, finally, we utilised prototypical network to obtain susceptibility maps. The results revealed that, under the influence of construction, the area of regions susceptible to collapse increased by 54.27% relative to the 2008 baseline by 2023, and is projected to reach 61.75% by 2030. Also, land use type and the density of linear projects significantly impacted model predictions and correlated with increased susceptibility levels. Our study offers a replicable template for other cities assessing the impact of construction on geological hazards.