Discussion On The Attribute Reduction Method In Inconsistent Decision Formal Contexts

Formal concept analysis is proposed by Professor Wille.R from German in1982, which is based on formal context to analyze formal concept. As an effective tool of knowledge acquisition and data analysis, the core of formal concept analysis is concept lattice. Formal context is considered as a kind of special information system, and it is a binary context made up of objects and attributes. At present, formal concept analysis has been studied extensively, and successfully applied to the area of as machine learning, software engineering and information acquisition, etc.Attribute reduction has been being one of the hot topics of this field since formal concept analysis is proposed. This paper considers how to conduct attribute reduction when decision formal context is inconsistent, at the same time, it considers how to apply the reduction method(approximate reduction) in electrical system fault diagnosis. In this paper, the main research achievements are as follows:(1) Attribute reduction is studied and approximate reduction is introduced when decision formal context is inconsistent. Firstly, some concepts are defined. For example, lower approximation coordination sets, decision reduction sets, and so on. Secondly, the discrimination methods of lower approximation coordination sets and decision reduction sets are proposed. Thirdly, approximation discernibility attribute sets and approximation discernibility attribute matrices are constructed. Finally the approximate reduction method is introduced, and it is demonstrated by experiments.(2) The approximate reduction method is applied to extract decision rules in power system fault diagnosis, so as to diagnose fault better. Firstly, decision formal context of fault diagnosis and the confidence of decision rules in fault diagnosis are defined. Secondly, fault decision formal context of some power system is reduced by the approximate reduction method. Finally the decision rules of fault diagnosis with their confidence are obtained. By this way, it helps the operator judge the possible sources(or regional) of the fault components based on decision rules in the event of a fault, with the aim of timely and accurately repaired fault.