Research On Multicriteria Classification Method On Dominance-Based Rough Set

Multicriteria classification refers to the classification of decision classes with priority by a set of preference-ordered conditional attributes,(also known as conditional criteria).In the existing studies,dominance-based rough set approach based on the dominance relationship has been successfully introduced into the multicriteria classification problems to express and explain the problems inconsistent with a dominance principle.In practice,attribute values with a hierarchical structure allow people to organize,view,and analyze data from different perspectives to adapt to changes in preferences.However,in the existing research,hierarchical multicriteria decision system of attribute reduction is not involved,incremental learning about approximation updating in the existing research is only restricted to complete multicriteria decision system,in view of this,this paper have studied the attributegeneralization reduct of complete multicriteria decision system,furthermore,the attributegeneralization reduct is extended to the imcomplete multicriteria decision system,and the corresponding approximation updating algorithm is given.Firstly,we introduce the attribute value taxonomy and cut into the complete hierarchical multicriteria decision system,and study the relationship between attribute reduction and attribute-generazation reduct under the finestlevel criterion values.It is proved that attribute reduction is a special attribute-generalization reduct.From this,we start off the criteria of the finest-layer attribute reduction of the attribute value taxonomy,making use of upwards ranking condition entropy for attribute-generalization reduct,designing a top-down search algorithm for attribute-generalization reduct in complete hierarchical multicriteria decision systems,and through the example illustrates in detail the whole process of the algorithm,experimental verification was given.The results show that the proposed algorithm for attribute-generalization reduct,which can objectively generalize attributes to maximize high levels while keep the same classification capability as the raw data.From that,we can accomplish better classification performance and compute smaller rule sets with better generalized knowledge.Secondly,attribute value taxonomy and cut are introduced into incomplete hierarchical multicriteria decision systems,studying the incremental updating of rough approximation approach between different granularity knowledge in incomplete hierarchical multicriteria decision systems,given the updating algorithms in dominating class and the decision upwards joint of cut refinement.Finally,the attribute-generalization reduct algorithm is extended to the incomplete case,and it is verified by an example.