Researched For Processing Approach Of Incomplete Information System Based On Rough Set Theory

Because the impact of various factors, the pending data are often incomplete in practical data mining problems. However, these missing data will often affect the subsequent data analysis. In this case, how to acquire knowledge from incomplete information system has become a crucial research topic recently. Rough set theory is an effective method in processing incomplete, imprecise and uncertain information. It is characterized by finding out nature characteristics and internal rules of of the given problems without needing any priori knowledge. Thus, incomplete information processing approaches based on rough set theory are studied systematically in this dissertation. The main content of this dissertation contains the following two aspects:First, the main methods of filling incomplete data currently are introduced and the merits and demerits of them are analyzed especially the ROUSTIDA algorithm. On the basis of ROUSTIDA algorithm, we propose an incomplete data filling algorithm based on attribute importance(IDFAAI). It is characterized by considering the impact that attributes importance for filling missing values in incomplete information system. Experiment show that compared with the ROUSTIDA algorithm, our algorithm has a higher filled rate and accuracy rate.Second, two kinds of attribute reduction algorithms of incomplete information system are introduced under the expansion rough set model based on tolerance relation. One is based on extended discernibility matrix and discernibility function, and the other is based on information entropy. After summing up the characteristics of above algorithms, combined with IDFAAI algorithm, we propose an attribute reduction algorithm based on tree (ARAT). The characteristics of the algorithm are able to get the complete reduction of decision table, and have a high computational efficiency. Finally, using the analysis of examples, the effectiveness of the algorithm is verified.