Abstract: In view of uncertain factors and time-location randomness in underground mine water inrush，the hazard identification and risk prediction of underground mine water inrush was studied by improved D-S (Dempster Shafer） Algorithm，so as to provide effective theoretical support to identify hazard as well as prevent water inrushes.In consideration of natural geological factors，more attention was paid to human factors in mine production management.With expert analysis，the BPA (Basic probability assignment）were governed to the factors which affect underground mine water-inrush.Based on the improved D-S Algorithm，the frame of discernment Ω was proposed which includes water-inrush，critical condition，no water-inrush and uncertain information.And the water-inrush integration decision-making model was established.The evidences of the model were aquifer water-bearing ratio，water-resisting layer thickness ratio，underground water pressure ratio，structure ratio，underground pressure ratio，exploration ratio，mining ratio and preventive management ratio.Finally，in the case of an internal typical underground mine，the results were broadly in line with the actual situation，which showed that the model was feasible and applicable.

Key words: Mine water inrushes, Improved D-S algorithm, Hazard identification, Water-inrush risk prediction