The Attribute Reduction In Real-valued Information Systems And Its Applications

Rough set theory proposed by the Poland mathematician Pawlak in1982, is a dataanalysis tool, it can be used effectively to deal with incomplete, incompatible and impreciseinformation, and discover the implication knowledge, reveal the potential rules. Reduction issignificant in data mining, also is one of the core issues of rough set theory. Researcherproposed many attribute reduction theory and algorithm for different information systems.With the real-value information system, this paper discusses rough set model and fuzzy roughset model based on tolerance relation, and study the attribute reduction approach inreal-valued information systems. Furthermore, we apply our theoretical results to the field ofradio signal recognition. The specific contents are as follows:1. With the help of sample mean and variance, the similarity measure between any twoobjects with respect to each attribute is given in the real-valued information system, and thenwe define a tolerance relation with a certain similar level, build a rough set model in thereal-valued information system, and give attribute reduction based on discernibility matrix.Furthermore, we discuss the distribution reduction of the consistent and inconsistentreal-valued decision table under the tolerance relation, and obtain the judgment theorems andthe specific method for reduction calculation.2. This paper is to discuss the fuzzy relative attribute reduction of the real-valuedinformation and real-valued decision table from the viewpoint of fuzzy rough theory. Firstly,based on the similarity measure between the objects in a real-valued information system, wedefine a fuzzy tolerance relation, and obtain approach to attribute reduction of the real-valuedinformation based fuzzy information quantity. Secondly, we define the fuzzy upperapproximation and lower approximation of decision classes. Also, by introducing fuzzyapproximation quality, we define the fuzzy significance and relative significance of conditionattributes, and propose a heuristic algorithm for computing the fuzzy relative reduction of thereal-valued decision table.3. With C-band radio signal monitoring as the application background, on the basis ofpretreatment of radio signal, we make the radio signal database into an abstract real-valuedinformation system. By using the reduction method which is based on the attributesignificance, we obtain the reduction features of the radio signal database.