Study on Knowledge Expression and Efficient Attribute Reduction Algorithm Based on Information Granule

Abstract

Data mining refers to extracting interesting (non-trivial, implicit, previously unknown and potentially useful) information or patterns from data in large databases. Noisy and inconsistent data are commonplace properties of large database and data warehouses. It is difficult when noisy and inconsistent data are mined by using classical rough set theory. In this paper, the concept of information granule is introduced. Then the knowledge possessing given confidence is described by using concept of information granule and the roughness and simpleness of knowledge is discussed by using extension of rough set theory. At last, the algorithm for attribute reduction based on information granule is presented. Experimental results show that the presented algorithm is good at enormous data production and effective to extract simplicity knowledge from noisy and inconsistent data with minimum confidence threshold.