Knowledge Discovery Research Of Fuzzy Information System

Rough set theory is a new mathematical tool of dealing with uncertain information, scientist Pawlak put forward this theory at 1982. At present it has developed to be an important research direction of artificial intelligent, it has very extensive latent applying background at the field of Data Mining.Pawlak rough set was established at the base of equivalent relation. For generalizing the application area of rough set theory, Pawlak rough set model is generalized to many models included fuzzy rough set model, probability rough set model, alterable precision rough set model and the rough set model under the generic binary relation etc. Fuzzy information system knowledge discovery based on fuzzy rough set model is researched in this thesis. Material work is done as follows:Firstly, in allusion to fuzzy information system, a similarity measure operator is defined to describe adjacence degree of objects on attributes. Fuzzy indiscernibility relation (R|~)_b induced by the similarity measure operator takes place of a series of a kind of more generalized fuzzy similarity relations. Based on the fuzzy indiscernibility relation, upper and lower approximation operators are defined in the generalized approximation space (U,(R|~)_b~α) and some proposition is discussed.Then, some theorems or proposition about some concepts such as fuzzy reduction and fuzzy core defined by Wang Xi-zhao are modifiedand improved for more completion and relations between fuzzy reduction and fuzzy relative reduction are discussed.Secondly, some concepts such as fuzzy relative reduction and fuzzy relative core between attributes are put forward and their basal properties are discussed. For the fuzzy decision information system with a decision attribute and a discrete domain, fuzzy relative discernibility matrix and fuzzy relative discernibility function are designed to obtain fuzzy relative reduction and fuzzy relative core.At last, based on fuzzy indiscernibility relation, dependence relations each other between attributes are discussed. Knowledge is discovered for the fuzzy decision information systems with more than a decision attribute, and a method obtaining rules of fuzzy information system is proposed. An applied effective example is given.