Study of Selective Ensemble Learning Methods Based on Support Vector Machine

AbstractDiversity among base classifiers is an important factor for improving in ensemble learning performance. In this paper, we choose support vector machine as base classifier and study four methods of selective ensemble learning which include hill-climbing, ensemble forward sequential selection, ensemble backward sequential selection and clustering selection. To measure the diversity among base classifiers in ensemble learning, the entropy E is used. The experimental results show that different diversity measure impacts on ensemble performance in some extent and first three selective strategies have similar generalization performance. Meanwhile, when using clustering selective strategy, selecting different number of clusters in this experiment also does not impact on the ensemble performance except some dataset