A Rule Extraction Method Based on Meta-information

Abstract

Rule extraction is an important research area of rough set theory. Many rule extraction methods, such as LEM2, are proposed. However, almost all these methods are on the assumption that they are dealing with a centralized dataset. A costly work of data integration is inevitable for these methods in case of distributed data environment. Meanwhile, meta-information is a compact description of information system or its sub-systems, and the cost of meta-information integration is much less than data integration. Moreover, since the volume of meta-information is much lower than the corresponding original dataset, the cost of operations on the meta-information is comparatively less. In order to take advantage of the meta-information mechanism, a minimal rule set extraction method is proposed in this paper on the basis of meta-information and the complexity of this method is much less than LEM2.