Abstract: For problem of difficult obtainment of abundance, comprehensive and accurate environmental information when data from a single sensor is as basis of underground safety assessment, an underground risk assessing method based on rough sets and D-S evidence theory was proposed, and analysis of specific example was given. The method makes first data fusion based on rough set and secondary data fusion based on D-S evidence theory to fuse data collected by multi-sensor, achieves integration of a variety of underground information, and optimizes existing risk assessment methods. Tests showed that the method can guarantee completeness and independence of forecasting information and realize dynamic prediction of mine safety trend.