Harnessing machine learning and multi-source data fusion for regional geohazard risk assessment of a rapidly urbanizing mountainous region

Rapid urbanization in mountainous regions often intensifies the conflict between engineering activities and the geological environment. in southeastern Sichuan Province, the study area is experiencing heightened geological disaster risks driven by its accelerated urbanization. To quantify this conflict, we developed an integrated hazard–vulnerability–risk assessment workflow. Hazard susceptibility was modeled using nine conditioning factors and three models—Information Value (IV), Analytic Hierarchy Process (AHP), and Random Forest (RF)—while vulnerability was mapped using a game-theory weighting method based on four socio-economic indicators. Results indicate that the RF model achieved the best discrimination (AUC = 0.936), significantly outperforming AHP (AUC = 0.798) and IV (AUC = 0.865). The RF-based high-hazard classes covered 36.79% of the area and captured 94.73% of historical sites. Population density and road density were the main contributors to vulnerability, with combined weights of 0.351 and 0.324, respectively, and the high and very-high vulnerability zones accounted for 21.13% of the study area. The high- and very-high-risk zones covered only 8.07% of the study area, including 4.72% high-risk and 3.35% very-high-risk areas, and were mainly concentrated along the riverside urbanization belt and areas with dense population and infrastructure. These results provide a basis for targeted monitoring, spatial planning, and geohazard risk management in rapidly urbanizing mountainous areas.