Abstract: In response to the increasingly prominent issue of road surface collapses driven by high-intensity underground space development in major and medium-sized cities in China, this study proposes a scientifically robust susceptibility assessment model to analyze spatial distribution patterns and support safe urban planning and disaster prevention. The aim is to mitigate the threats posed by road collapses to urban public safety and infrastructure, ensuring the safety of residents' lives and property. Based on 38 road collapse incidents recorded in Hangzhou from 2008 to 2024, twelve evaluation indicators were selected to construct a localized susceptibility index system. Feature importance ranking and collinearity analysis were conducted to refine the indicators. To address the sample imbalance issue, a negative sample selection method based on grid cell size was adopted, and the Synthetic Minority Oversampling Technique (SMOTE) was applied to balance the positive and negative samples. On this basis, a random forest model was then used to compute a road collapse susceptibility index for Hangzhou. Finally, the natural breaks classification method is used to divide the study area into high, medium, and low susceptibility zones —completing the susceptibility evaluation of road collapses in the study area. The results indicate that road collapse susceptibility in Hangzhou exhibits a spatially decreasing pattern radiating outward from the central plains of Gongshu, Shangcheng, Binjiang, and Xihu districts. Specifically, high-susceptibility zones account for 12.36% (325.87 km²) of the area, medium-susceptibility zones for 30.45% (802.52 km²), and low-susceptibility zones for 57.19% (1,507.56 km²). This research reveals the spatial distribution patterns of road collapse susceptibility in Hangzhou and offers scientific guidance for the prediction and prevention of road collapse susceptibility in Hangzhou.