Rough Set Models Based On Dominance Relation And Maximal Consistent Blocks In Incomplete Information Systems

Missing attribute value is one of the main characteristics of complex data,and the data table composed of such data is called an incomplete information system.How to discover knowledge from incomplete information systems has become a hotspot and difficulty in the field of data analysis.Currently,extended rough set models have been widely concerned by scholars,and progress has been made in acquiring knowledge from incomplete information systems.Aiming at generalized incomplete information systems,this paper has constructed several extended rough set models and decision-theoretic rough sets,and explored their mathematical properties and relevant relationships.The main contents include:To explore incomplete ordered information systems,local double relative quantitative decision-theoretic rough set(LDrq-DTRS)models are coined from the viewpoint of double relative quantitative information through introducing the local probabilistic graded rough set and local decision-theoretic rough set based on dominance relation.The corresponding decision rules are given,and the mathematical properties of the proposed models are investigated.In addition,the relations between the decision regions of two LDrq-DTRS models under different parameter relationships are discussed,and the inner relations between a pair of LDrq-DTRS models and the dominance-based rough set model are studied.As regards generalized incomplete information systems,this paper combines the graded rough set with decision-theoretic rough set to put forward two double quantitative decision-theoretic rough set models based on maximum consistent blocks(MCB-Dq-DTRS)from the perspectives of absolute quantitative and relative quantitative information.On this basis,the mathematical properties of the proposed rough set models are discussed.Additionally,the inner relations and the fault tolerance capability between two MCB-Dq-DTRS models and the MCBRS model are analyzed under different parameter relationships.