Consistent approximate representation space for propertyoriented concept lattices
Junyu Li,Xia Wang,Qingfeng Liu
DOI: https://doi.org/10.13232/j.cnki.jnju.2017.02.015
2017-01-01
Abstract:Formal concept analysis is an effective tool for data analysis,knowledge discovery and information management.This paper mainly studies a unified structure for a formal context,a consistent formal decision context as well as an inconsistent formal decision context based on a property oriented concept lattice,and proposes a new approach to attribute reduction for property oriented concept lattices of different kinds of formal contexts.Using equivalence relations defined on the object power set,an approximate representation space is first introduced for a formal context,which is a quadruple including an object set,an attribute set,a family of equivalence relations on the object power set and an equivalence relation on the object power set.And then the notion of consistency of the approximate representation space is defined.Moreover,consistent approximate representation spaces are constructed corresponding to a formal context,a consistent formal decision context and an inconsistent formal decision context,respectively.In particular,four types of approximate representation spaces are defined for an inconsistent formal decision context,which are distribution approximate representation space,maximum approximate representation space,lower approximate representation space and upper approximate representation space.It is shown that a formal context,a consistent formal decision context and an inconsistent formal decision context are all special cases of consistent approximate representation spaces.Finally,a notion of attribute reduction is defined for consistent approximate representation spaces,and an approach to attribute reduction to calculate all attribute reducts is presented by means of discernibility matrices.It is proved that attribute reduction of six types of consistent approximate representation spaces are just corresponding to attribute reduction of a formal context,a consistent formal decision context and an inconsistent formal decision context.The result indicates that attribute reduction of consistent approximate representation space can be regarded as the unified approach to attribute reduction of different kinds of formal contexts,which can reduce the complexities of attribute reduction produced by differences of formal contexts.