Plenary lecture 6: method for classification in interval-valued information systems
Amaury A. Caballero · 2010
Due to its importance in a number of fields, the problem of classification or classes discrimination in information systems has been approached by multiple authors, and various methods to address this issue have been developed. The combination of rough sets and fuzzy logic for classification is a widely adopted method. Rough set theory helps in minimizing the number of attributes that influence the selection and fuzzy logic permits to discriminate when there is more than one possible solution for the same attributes and intervals. Neural networks and information entropy have also been used to discriminate. When information is diffuse and the number of obtained values for each attribute is large, such is the number of rules for any type of solution method. Due to this fact, interval-valued information systems have been proposed by several authors, in which an interval of values is defined for each attribute, moving from the minimum to the maximum obtained values in the database or using the standard deviation from the original data to define the minimum and maximum values. Differently from other works, the concept of information measure is used in this paper, together with a fuzzy logic discrimination tool. Using these concepts, an attribute reduction is initially obtained and then fuzzy logic is applied for discriminating among the possible solutions. The method results simpler than others, and as accurate as the methods usually employed.