Research Of Decision Implication On Attribute Granulating

Formal concept analysis(FCA)is a method of concept analysis based on formal context.Formal context shows binary relationships between objects and attributes,based on which FCA constructs concept lattice,visually analyzes logical relationship between concepts and mines hidden knowledge.Decision context divides the set of attributes in formal context into a set of condition attributes and a set of decision attributes.Decision implication,as an IF-THEN rule,reflects the strong correlation between condition attributes and decision attributes.In the existing knowledge representation of formal concept analysis,decision implication has strong knowledge representation ability.Granulation is a research method of granular computing.Granulation can be formed by refinement and generalization.Attribute granulation is helpful to extract the knowledge of different attribute granularity layers in data,thus with important application significance in knowledge representation and reasoning.In addition,decision implication canonical basis(DICB)is the most compact set of decision implications,which can efficiently represent all the knowledge in decision context.Therefore,this paper mainly studies the generation of decision implication after granulation by studying the generation of DICB after granulation.Specially,we study condition attributes granulation and decision attributes granulation respectively,analyze the changes of DICB under different conditions,and present corresponding generation method of DICB.In practical applications,users can generate DICB according to this method,and then obtain all the decision knowledge of granulation context according to these methods.The research results of this paper are as follows:(1)The generation method of decision implication is studied after condition attribute is refined.First,the generation of decision implication can be completed by studying the generation of DICB.Since the decision premise is the basis for generating DICB,the changes of DICB can be analyzed by classifying decision premises into different types and studying the changes of the types.In order to simplify,we study the changes of decision premise in two stages: deleting coarse-grained condition attributes and adding fine-grained condition attributes.Based on this,we design a true premise based incremental method for the generation of DICB.Finally,the effectiveness of the proposed method is verified through experiments,and its advantages over the generation method based on true premise are analyzed.(2)The generation method of decision implication is studied after decision attribute is generalized.In order to obtain more coarse-grained decision information,we generalize decision attributes.Similar to the refinement of condition attributes,we divide the generalization process into two steps: deleting fine-grained decision attributes and adding coarse-grained decision attributes,and analyze the changes of decision premises in each step.It is found that in the generalization context,it is sufficient to generate the true premises of coarse-grained decision attributes.Accordingly,an iterative generation method of DICB for generalization context is proposed.