A Novel Attribute Reduction Algorithm Based on Rough Set and Information Entropy Theory

The incompleteness of measurement approach of importance of attribute that is based on condition entropy is analyzed and proved through example. After the information entropy of element in positive region is introduced in the measurement of importance of attribute, both a novel measurement approach of importance of attribute and a novel measurement approach of importance of single attribute relative to attribute set are put forward. Based on above ideas, a heuristic attribute reduction algorithm is constructed by adopting SGF*(a, A, D) as heuristic information. Finally, the feasibility of the measurement approach of importance of attribute and the validity of the heuristic reduction algorithm are demonstrated by some classical databases in the UCI repository.