A Study And Application Of Rough Set Theory In Knowledge Acquisition Incomplete Information System

Rough set theory is introduced by Professor Z.Pawlak in Warsaw universityof technology, Poland, which is a research method of organization and analyzeof incomplete, inaccurate or vague knowledge. Since it has been introduced, rough sets have been used widely in many fields, such as model recognition, datamining and decision analyze. Rough set theory provides new method and toolsfor data mining. The applications of rough set theory in data mining mainly i-nclude two aspects: attribute r-eduction and rules extraction. According attributereduction and decreasing dime-nsions, rules would be acquired adapted for deci-sions. Classic rough set theory is based on complete data sets, though data fro-m practice are usually incomplete. So how to acquire knowledge from incompl-ete information system has been a crucial research topic recently.Firstly, this paper analyses methods of attribute reduction and rules extractionin complete information system, and gives correlative algorithms and examples.Then based on extension of classical rough set theory, we discuss methods ofattribute reduction and rules extraction in incomplete information system. Meth-od of attribute reduction methods based tolerance matrix, attribute reduction met-hod based core and attribute reduction method based information entropy are analyzed and compared, and correlative extension properties are proofed. Correspo-nding algorithm are provided. The next part is about heuristic rules of attributereduction methods in incomplete information system, and introduces a new heur-istic rule based on tolerance matrix and principle component analyze and corres-ponding attribute reduction algorithm. Lastly, rules extraction method based tole-rance relation and discernibility matrix.Lastly, a simple application of rough set theory in care and management syst-em of roads and bridges is provided. According data collected from actual circ-umstance, complete decision table and incomplete decision table which is builtaccordi-ng to 10% missing data formed randomly are established respectively, a-nd some basic decisive rules which are used for care and management of a bri-dge are acquired, according to corresponding theories of rough set in completeinformation system and incomplete information system.