A novel knowledge reduction method based on rank correlation analysis

Dominance-based rough set approach has recently become a routine method to deal with preference-ordered data, and knowledge reduction method based on rough set theory has been proposed. However, the results obtained are usually short of statistical significance. In this paper, non- parametric methods in statistics are introduced to analyze ordered information systems and ordered decision tables. Spearman and Kendall rank correlation coefficient are respectively used as new measures of attribute sets correlation. Based on these measures, a new method of knowledge reduction of the ordered information systems and the ordered decision tables using nonparametric rank statistics is presented. It can be proved that there are some relationships between the rough set theory and the nonparametric statistical methods. The numerical experiments show that the approach proposed is feasible, and it can provide a statistical evidence for rough set method.