Abstract: A rough rules mining approach based on variable precision rough set theory is proposed in this paper.With uncertain measurement of rough rules set,relative attribute reduction is obtained by applying GA,and then rough rules set is deduced under the threshold values.The effect of threshold values on the rules set is discussed in this paper.The rules obtained by this approach have a certain noise tolerance,more precision and better overlay.So the accuracy of forecasting and classifying are assured.The practical results show that the approach is effective in solving knowledge mining problems.