Abstract: Attribute reduction is important research content of rough set theory.Attribute reduction based on information entropy is an effective method of knowledge reduction.In practical application,the acquired information system is usually not complete.To solve this problem,we defined a new knowledge entropy based on the relationship between the attribute subset redu and CAttr-redu,and proposed a new incomplete information system attribute reduction algorithm (newS algorithm) applying approximate fuzzy entropy.Finally,simulation experiment was carried out on 6 data sets in ROSE and UCI data.The experimental results show that the newS algorithm is feasible,and has higher efficiency compared with other algorithms under the same reduction effect.

Key words: Incomplete information system,Fuzzy entropy,Attribute reduction