Combination Of Frequency Reduction And Dynamic Reduction Methods For The Classification Of Inconsistent Decision Tables

Rough set theory is a new mathematic tool for processing uncertainty information after probability theory, fuzzy set theory and D-S theory of evidence. It is very effective for analyzing imprecise, incomplete or inconsistent information, and discovering underlying knowledge or characteristic. Because of Its good performance in uncertain area, Rough set theory has many successful applications in machine learning, knowledge extracting, decision analysis, data mining, expert system, decision support system, inductive reasoning or scheme recognizing, etc.Because the decision rules computed from inconsistent tables is uncertain, we will get many different result and cannot gain the certain knowledge about a given unknown object when using these rules to classify the unknown object. However, decision tables in general is inconsistent tables because of definition, criterion and operation in data collecting. So, How to reduce and eliminate the negative affection of inconsistent tables has been became a very important research subject.In this paper we propose a method for classification for inconsisitent tables by combining frequency reducts with dynamic reducts on the basis of rough set theory. On the one hand, frequency reducts can suppress and eliminate the noise and the negative affection of inconsistent tables the furthest, and make the most of the information contained in inconsistent tables. On the other hand, dynamic reducts are in some sense the most stable reducts of a given decision tables and are very appropriate to classify unknown object, because they are the most frequently appearing reducts in subtables created by random samples of a given decision table. So combining frequency reducts with dynamic reducts and making the best of good character of two methods can obtain better quality and stability of classification.