The Attribute Reduction And Rules Acquisition Based On Set-valued Information Systems And Its Applications

.Rough set theory is an approach of data analysis dealing with uncertain and incomplete information. With the developing of Internet, the quantity and complexity of business data are increasing continuously, and the uncertain of information is more and more remarkable. The set-valued information system is a representation of incomplete information, it is reasonable to deal with many problems in actual life. In this paper, under the framework of set-valued information system, the extended model of rough set based on tolerance relation is discussed. By the concept of heuristic algorithm on attribute reduction, the method on attribute relative reduction ofλ-lower approximation is presented. These researches are applied to the database of radio signal, and used to reduct the signal character and acquire the classification rules. It is the foundation for intelligent analysis of radio signal. The main content of this paper can be summarized as follows:1,In a set-valued information system, it is generalized for quantifiers which is used to define the model of rough set and based on maximal tolerance classes. Associated with the parameterλ∈[0,1], and replaced existential quantifier"?"and universal quantifier"?"with fuzzy quantifiers, the definition onλ-approximation operators, which based on the maximal tolerance classes, is proposed in set-valued information systems. Meanwhile,λ-upper andλ-lower approximation sets are given, and the properties ofλ-approximation operators is discussed.2,Based on the concept of heuristic algorithm on attribute significance, the method on attribute relative reduction of set-valued information systems is proposed by constructing the membership function ofλ-lower approximation set, and defining the importance degree of condition attributes for decision attribute. Further, associated with the maximal tolerance class, the definition on certain and uncertain rules, which is related toλ-level, is proposed in set-valued information systems.3,Under the background of intelligent analysis of radio signals, the database of continuous values is transformed to the set-valued information system, by fuzzy clustering the feature values and discretizing the signal database. The reducted features are obtained by the method on attribute relative reduction ofλ-lower approximation, and furthmore, the classification rules are also obtained to determine the type of radio signals.