Reduction And Rule Acquisition In Fuzzy Decision System Based On Neighborhood Relationship

With the development of the economic society, the systems people have faced are constantly increasingly complex; the things we depict and describe have different degrees of uncertainty and ambiguity. Representing things using fuzzy concept often appears in real life, for example:during a medical treatment, we only need to determine whether a patient is myopia and the degree of his myopia. The degree for an object belonging to a fuzzy concept can be expressed by fuzzy membership functions named fuzzy membership. When a decision attribute of a decision information system is a fuzzy concept, we name it as a fuzzy decision information system, called fuzzy decision system. Fuzzy decision systems are widely used in scientific research, construction, satellite remote sensing, and medical care.This thesis uses the basic idea of rough set theory and granular to make a research for knowledge acquisition methods of fuzzy decision system, which mainly includes three aspects as follows:(1) Defining the concept of fuzzy decision system. By a given distance metric, we, on the domain space U of a continuous range of fuzzy decision system, construct neighborhood and use it to make a partition. And also we discuss the properties of the neighborhood and extend it to a nominal type of fuzzy decision systems.(2)About the fuzzy decision systems, we propose rough set model, which is based on equivalence relation, tolerance relation, neighborhood relation, and discuss the links among the three. We define the concept of upper and lower consistent attribute reduction of fuzzy decision system, and establish discernibility matrix using parameters (α,β), and through the discernibility matrix of fuzzy decision system, we define upper and lower consistent attribute reduction, (α,β) distribution of consistent reduction, and propose the attribute reduction methods under the equivalence relation, tolerance relation ,and neighborhood relation, and design an attribute reduction algorithm of fuzzy decision system based on general neighborhood relation.(3)Based on reduction, we define decision rules, and also give the computation method of local confidence level about the decision rules, and propose that the value of 8 is more critical to the local confidence level. Taking the premise of ensuring local confidence level, by the neighborhood and decision-making value we merge the decision rules to get simple and effective rules to achieve the predictive purpose.In this thesis, based on rough set theory and granular idea, we make a preliminary exploration and research for fuzzy decision systems; especially present a knowledge acquisition method for fuzzy decision systems with continuous values of condition attributes. The methods of attribute reduction and rule extraction methods in the paper are validated by experiments, and prove that the granular idea can be effectively applied in fuzzy decision systems with continuous values of condition attributes, which provide a new method for knowledge acquisition of fuzzy decision systems with continuous values of condition attributes.