Research Of Model And Attribute Reduction Algorithm For Set-Valued Information System Based On Rough Set

With the rapid development of information technology, information systems have accumulated large amounts of data. So an efficient toll for processing information is in urgent need which can extract and mining those implied but unknown and valuable knowledge which can be used for people in mass data. In1982, the theory of rough set was proposed as a new mathematical tool to deal with fuzzy and uncertain data of which the important feature is to export the concept classification rules directly form the given description and does not require any prior knowledge in maintaining the classification unchanged cases. And after30years of research, it has made tremendous progress and become one of the most widely used theories.The model and attribute reduction are the two hot spots in the study of rough set theory today. As the classical rough set model and the traditional attribute reduction algorithms have failed to meet the needs of reality, many scholars have done with varying degrees of research both in model building and algorithm improving areas. But it still faces enormous challenges to put forward a more reasonable classification model and an efficient attribute reduction algorithm in dealing with mass data. This paper has made thorough studies in two aspects of model building and attribute reduction which are described as follows:(1) In views of the defects of previous proposed model, an improved rough set model called distance relationship model which is a binary relationship model based on distance is put forward on the basis of κ-degree limited compatibility relations. Because it learns the advantages that the threshold k is flexible and the similar relations are symmetrical in κ-degree limited compatibility relations so it can solve the division problem of studied objects in a single property. And the related properties of the model are studied on the basis of the three equivalence relation operators. Demonstrated, the loose degree of the model classification is between the compatible relations and dominance relations and the classification is better than the similar relations and κ-degree limited compatibility relations.(2) In this paper, an attribute reduction method based on β similarity is studied due to that set-valued information system contains a large number of data. Under the variable precision relation, β is used to constraint the similarity degree between objects so we can adjust the divided particle size of the class and the complexity of the difference matrix according to the change of β to carry on the attribute reduction. The algorithm is proved to be effective and feasible both in time complexity and space complexity through practical examples.(3) A knowledge distance-based attribute reduction algorithm is put forward after studying the set-valued information system and the nature of knowledge distance. The algorithm firstly uses the knowledge distance to describe the gap between the knowledge and then measures the classification results and knowledge granularity of the set-valued information system model and finally it effectively judges the loose degree of the set-valued information system model and the importance of attributes depending on the nature of the knowledge-based distance. In addition, it shows that the algorithm reduces the time and space complexity thereby improving the operating efficiency through theoretical analysis and experimental results.