Rough Set Theory-based Multi-class Decision Attribute Reduction Algorithm and Its Application

Rough set theory is an effective tool in dealing with vague and uncertainty information, attribute reduction is one of its important concept. Many attribute reduction algorithms have been proposed, but they are more suitable for two classes problem. For multi-class decision attributes problem, a new attribute reduction algorithm based on discernibility matrix is proposed in the paper, it makes great use of the advantage of decision attribute´s class information. In addition, we may draw an important conclusion that attribute reduction connects with class information in multi-class decision system, that is to say there will be deferent reduction results between deferent classes. The proposed algorithm can effectively reduce the computational complexity and increase reduction efficiency. Finally it is applied to diesel engine fault diagnosis, diagnosis result shows its feasibility and validity