Research Of Attribute Reduct Algorithm Based On Rough Set

Rough Set theory is a new mathematical tool which can tackle ambiquity and uncertainty developed from the 1970s. It is an important method of intellective information transaction, which based of non-distinguish and technoledge reduction. At present, it's one of the important problems of calculating high effictive, shortcut attribute reduction algorithms, and the high effictive reduction algorithms have important application significance in information system analysis and data mining.This paper is based on rough set theory, due to there is consistence and inconsistence in decision rule of information system, which results in consistence and inconsistence of decision table. However most of attribute reduction algorithms of decision table need to judge whether decision table is consistent or inconsistent before progressing attribute reduction. Aiming at consistent or inconsistent of decision table, researches representative attribute reduction algorithms respectively. Consistent decision table includes an algorithm of attribute reduction based on dependence and importance and an algorithm of attribute reduction based on generalized information table; while an algorithm of attribute reduction based on inclusion degree in the inconsistent of decision table. It learned that above attribute reduction algorithms may only reduce consistent or inconsistent decision table respectively, wasting of most time, and they had not universality. In order to takle that, a proposed algorithm of attribute reduction for decision table based on entropy is pointed. The experiment result and above ones are accordant through analyzing. The new algorithm need not judge whether decision table is consistent or inconsistent before progressing attribute reduction for decision tables. When these definitions are used to reduce inconsistent decision table, reduction results could be inconsistent, it solves this problem and advances efficiency. It has universality and reaches anticipative effect.