Abstract: To improve the efficiency of attribute reduction,a rapid reduction algorithm based on conditional information quantity is proposed.The concepts of information quantity based on partition and conditional information quantity based on partition are defined,and the theorems about the monotone of attribute significance、invariance of conditional information quantity based on partition are proven;Compared with other algorithms,this algorithm reduces the searching space of attributes and samples in each step.Experimental results showed that the rapid reduction algorithm is more efficient than the existing algorithms.