A new Ensemble Learning Method Based on Pairwise Constraints and Subset Selection.

It is well known that in order to get a strong ensemble, obtaining more accurate and high diverse component classifiers is very important. Most existing ensemble methods are based on the two conditions. In this paper, considering the two conditions as well, a novel ensemble approach based on pairwise constraints and subset selection are presented. Experiments on 7 UCI datasets are carried out and the results indicate our proposed method performs better than Bagging. [ABSTRACT FROM AUTHOR]