Inconsistent decision reduction of object oriented concept lattices
Junyu Li,Xia Wang
DOI: https://doi.org/10.13232/j.cnki.jnju.2016.05.011
2016-01-01
Abstract:In practice ,most formal decision contexts are inconsistent because of various factors such as prediction ca‐pability ,noise in data ,and so on .Due to inconsistency ,it is more complex and difficult to extract useful information from inconsistent formal decision contexts .Since attribute reduction of concept lattices makes knowledge representation of formal decision contexts more succinct ,knowledge hiding in formal decision contexts clearer ,and adaptability of rule sets for formal decision contexts better ,it is necessary to research attribute reduction of concept lattices in the case of inconsistent formal decision contexts .The purpose of this paper is to investigate notions and methods of attribute reduction of object oriented concept lattices in the case of inconsistent decision formal contexts . Based on an equivalent relation defined on the object power set ,two notions of attribute reduction of an object oriented concept lattice are presented for an inconsistent decision formal context ,which are decision attribute reduct and maximum decision attribute reduct .It is shown that the distribution consistent set preserves the degrees in which the conditional equivalent class belongs to each decision equivalent class ,and the maximum distribution consistent set preserves all maximum decision equivalent classes .Then relations between decision attribute consistent set and maximum decision attribute consistent are discussed in detail .And a distribution consistent set must be a maximum distribution consistent set ,but the converse is not necessarily true .Finally ,two notions of discernible matrix are introduced into an object oriented concept lattice ,which are decision discernible matrix and maximum decision discernible matrix .By using decision discernible matrix and maximum decision discernible matrix ,judgement theorems of decision attribute reduct and maximum decision attribute reduct are obtained respectively ,and a method of discernible matrix is proposed to obtain all decision attribute reducts and maximum decision attribute reducts of the object oriented concept lattice .