Attribute Reduction For Sequential Three-way Decisions Under Dominance-Equivalence Relations

Sequential three-way decision is an effective way to solve problems under multiple levels granularity.However,most of current sequential three-way decision models are limited to information systems of symbolic attributes,which cannot process continuous and ordinal values effectively,and will cause a certain degree loss of information.On the other hand,dominance-equivalence relation based rough set approach can be used to handle classification problems with preference ordered conditional attributes,extract related information,and then approximate target concepts and finally form decision-making knowledge.The weakness of the traditional dominance relation-based rough sets model is very time consuming for knowledge reduction and extraction.Besides,since many real-world information systems are dynamic,attribute reductions need to be often updated as the most important knowledge in decision making.To tackle the previous mentioned issues,this thesis mainly conduct the following research work:In this thesis,we consider multi-criteria classification problems which refer to a type of classification problems with ordinal conditional attributes.The concept of dominance-equivalence relation is used to describe information systems of this kind of problems.Then we apply the idea of sequential three-way decisions to the dominance relation-based rough sets models.A new attribute reduction method is proposed based on sequential three-way decisions and the defined attribute importance measure,and then the processing of information systems with ordinal attributes is accelerated.As a result,the efficiency of knowledge reduction is improved through multiple granularity representations and relationships.Another contribution of thesis work is: In order to deal with the dynamic information system with preference relations and provide an efficient method for updating attribute reductions for multi-criterion decision-making problems,we established an efficient knowledge updating method based on sequential three-way decisions under dominance-equivalence relations.Multi-granules are combined to form dynamic granular sequence,the attribute reducts are updated through reusing the original information when the object set or attribute set are changed,thus saving the computing time and storage of attribute reduction process.Several UCI data sets are selected for experiments.The results show thatthe proposed method can obviously reduce the time consumption at the same time guarantee the quality of the attribute reduction.