关键词: 粗糙集, 属性重要度, 故障诊断, 属性约简, 决策

Abstract: The attribute significance degree based on the knowledge base of rough sets shows the classification change after removing one or more attributes. For the inadequacy of the method to determine the attribute significance degree, the direct and indirect effects of the condition attributes acted on the decision attributes are considered, and the method to determine the attribute significance degree, which is based on the attribute dependency degree in rough sets, is improved. Moreover, after comparing the difference between the original and the improved method, the meaning of the later is discussed, and theorems show that the improved method is more credible. Finally, the simulation on the attribute significance degree of mechanical faults shows that the attribute significance degrees are more distinguishable and suitable for the attribute reduction, which are beneficial to decision-makers’ quick judgments.

Key words: rough set, attribute significance degree, fault diagnosis, attribute reduction, decision