Attribute reduction of multi-valued information system based on conditional information entropy

In this paper, the concepts of information entropy and conditional information entropy in multi-valued information system are proposed from the viewpoint of information theory. Then, an algorithm for attribute reduction is presented in which a new kind of attribute significance is used as the heuristic information to reduce the searching space during the reduction of knowledge. The experimental results show that the valid algorithm can find the minimal reduction in multi-valued information system.