Heuristic Reduction Algorithm Based on Pairwise Positive Region

To guarantee the optimal reduct set, a heuristic reduction algorithm is proposed, which considers the distinguishing information between the members of each pair decision classes. Firstly the pairwise positive region is defined, based on which the pairwiae significance measure is calculated between the members of each pair classes. Finally the weighted pairwise significance of attribute is used as the attribute reduction criterion, which indicates the necessity of attributes very well. By introducing the noise tolerance factoc, the new algorithm can tolerate noise to some extent. Experimental results show the advantages of our novel heuristic reduction algorithm over the traditional attribute dependency based algorithm.