期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2017
卷号：17
期号：3
页码：223-227
出版社：International Journal of Computer Science and Network Security
摘要：Voting based Extreme Ensemble is a majority voting based ensemble of Extreme Learning Machines. An ensemble may contain some highly correlated classifiers. Ensemble pruning is used to remove these redundant classifiers, reduce the size of ensemble. It may also increase the accuracy of ensembles by selecting subsets of classifiers that, when combined, can perform better than the whole ensemble. This paper proposes a gain based ensemble pruning technique that adds classifiers based on their diversity and contribution towards ensemble. This algorithm reduces time complexity by reducing the size of pruning set which is done by eliminating training instances that are present far away from the decision boundary. These are the instances which are classified correctly by majority of classifiers with high confidence. Results show that the proposed algorithm works equally well or even better in some cases than Voting Based Extreme Learning Machine in terms of accuracy.
