Attribute Reduction Of Preference Linguistic-Valued Decision Information Systems

With the advent of the big data era,the diversity of data types has attracted more and more attention.To better describe the sample,there are many preference linguistic-valued information systems.Linguistic-valued data is natural language or artificial language data,which applied to describe some incomplete or vague complex phenomena.As a very popular mathematical tool to deal with incomplete and vague knowledge,rough set theory is proposed by Pawlak.The Pawlak's rough set model is constructed on equivalence relations.However,these relations are viewed as one of the main limitations when employing the model to some complex decision tasks.For dealing with preference linguistic-valued decision information systems,we propose a dominance-based fuzzy rough set model which is expanding of rough set model.This paper studies the reduction of preference linguistic-valued decision information systems.Firstly,we introduce the basic conception of Pawlak rough set theory,fuzzy logical and dominance-based rough set model.At the same time,we introduce some classical methods for the reduction of one information system,such as discernibility matrix and the significance of the attribute base on the independence.Secondly,for studying preference linguistic-valued decision information systems,we propose that every linguistic-valued data can be taken as one trapezoidal fuzzy number.we construct a dominance relation to measure the degree of one linguistic-value data dominant the other one.At the same time,we proposed a new way of aggregation to put two or more relation to one.In this way,we can relate every attribute with the set of all attributes.Based on this dominance relation,we propose the fuzzy rough set model and analyses the property.Based on the fuzzy preference relation,we propose a new dominance discernibility matrix to distinguish the linguistic-valued data.This matrix is asymmetrical.We construct a discernibility function base on logical operation,with which we can get all reduction of one information systems.To reduce the complexity,this paper proposes a measurement of the attribute significance based on information entropy.Based on this attribute significance,we use heuristic method to find the reduction of the preference linguistic-valued information systems.Simulated experiments verify efficiency and feasibility of the proposed algorithm.