Abstract: In view of the classical rough sets model can only dispose the complete information system,a variable precision rough sets model is introduced to dispose the incomplete information system.Firstly,the definition of(α,τ) limited similarity relation and its some interrelated characters are presented,and a variable precision rough sets model based on(α,τ) limited similarity relation is introduced.Secondly,the knowledge reduction algorithm based on attribute significance and the quality of approximate classification is proposed.Finally,an example is given to illustrate the validity of the proposed algorithm.