Study On Attribute Reduction Algorithms Based On Rough Sets

Rough Set theory is a new mathematical tool which can tackle ambiguity and uncertainty problems. It can deal with imprecise, uncertain, incomplete information and knowledge by quantitative analysis and processing. Attribute reduction, as a hot topic research area, is one of the core issues of rough set theory.At present, during the existing attribute reduction algorithms, applied to a complete and consistent information system, the efficiency and the completeness of attribute reduction algorithms have not been well resolved; applied to inconsistent information systems and incomplete information systems, approaches to attribution reduction are mainly based on discernable attribute matrixes, which are very time-consuming. In this paper, after studying a large number of relevant documentation, a relatively heuristic attribute reduction algorithm is presented. In this algorithm, for inconsistent information systems, using the conversion algorithm, three approaches are proposed to convert the inconsistent information systems into three different types of the consistent information systems, and a comparation and analysis of them is shown. For incomplete information systems, an extended rough set model is proposed to deal with the missing value which is called limited tolerance relation model to instead of non-distinguish relation. Finally, t the completeness of the algorithm is theoretically proved, and uses the UCI machine learning database for testing. Test results show that the attribute reduction algorithm, described in this paper, not only can deal with incomplete and inconsistent information, but also the efficiency has improved.