With　the　enhancement　of　extreme　rainfall,　urban　flooding　has　become　a　pressing　issue　for　high-density　cities.　However,　most　existing　studies　focused　on　the　physical　causes　of　flooding　and　associated　environmental　factors,　neglecting　multidimensional　factors　including　socioeconomic　and　spatial　elements.　Focusing　on　the　Guangdong-Hong　Kong-Macao　Greater　Bay　Area　(GBA),　this　study　proposes　an　innovative　analytical　framework　that　integrates　a　multidimensional　Hazard–Exposure–Vulnerability　(HEV)　assessment　system,　applies　the　Dagum　coefficient　to　evaluate　disparities　in　flood　protection　resource　allocation,　and　leverages　a　LightGBM–SHAP　model　to　quantify　the　synergistic　and　suppressive　effects　of　key　drivers　under　different　risk　scenarios.　The　results　indicated　that:　(1)　The　Bayesian-optimized　LightGBM　model　demonstrated　outstanding　accuracy　and　robustness　across　four　flood　scenarios,　achieving　an　average　R²　of　0.925　and　RMSE　of　0.005;　(2)　Flood　risk　followed　a　coastal-to-inland　gradient,　with　high-risk　areas　concentrated　in　densely　populated　urban　clusters　and　along　river　networks;　(3)　The　distribution　of　safety　resources　was　highly　uneven,　with　fragmented　planning　and　isolated　infrastructure　worsening　mitigation　inefficiencies　in　newly　developed　districts;　(4)　The　primary　flood　risk　drove　shift　from　topography　and　physical　attributes　to　hydrological　and　ecological　variables　at　higher　risk　levels;　(5)　Impervious　surface　proportion　(ISP)　and　fractional　vegetation　cover　(FVC)　were　identified　as　critical　determinants—ISP　contributed　35.15%　to　general　risk,　while　FVC　values　above　0.4　significantly　mitigated　flood　impacts.　By　integrating　risk　assessment　with　equality　analysis,　this　study　offers　practical　insights　for　improving　climate-adaptive　urban　planning　strategies.　©　2025　Elsevier　Ltd