Fuzzy positive region based attribute reduction in decision tables

Abstract

Traditional rough set based attribute reduction methods has performed on the decision tables with real value attribute domain needs to be discretized data. The discretized data can be lost information which will affect the quality of data classification. To overcome this drawback, attribute reduction performs directly on the decision table with real value attribute according to fuzzy rough set approach has proved effective. In this paper,
we propose two attribute reduction methods using fuzzy positive region based on fuzzy partition and fuzzy similarity relation. Analyzing and evaluating for each method which concludes the method using fuzzy similarity relation has practical application.