Attribute reduction of preference linguistic-valued decision information systems

Abstract

Dominance-based linguistic-valued information systems have attracted much attention in practical applications in which more information can be provided and employed than real-valued information systems. This paper proposes an approach to attribute reduction of dominance-based linguistic-valued decision information systems. All linguistic-values are represented by trapezoidal fuzzy numbers and a new fuzzy dominance relation is constructed to characterize the degree of one fuzzy number dominating another. Based on the presented fuzzy dominance relation, a model of dominance fuzzy rough set is developed and its properties are investigated. A way to measure attributes significance is established by introducing new fuzzy entropy. Comparative experiments are conducted to verify the effectiveness of the proposed model.