An Integrated Pruning Criterion for Ensemble Learning Based on Classification Accuracy and Diversity

Bin Fu, Zhihai Wang, Rong Pan, Guandong Xu, Peter Dolog

Abstract: Ensemble pruning is an important issue in the field of ensemble learning. Diversity is a key criterion to determine how the pruning process has been done and measure what result has been derived. However, there is few formal definitions of diversity yet. Hence, three important factors that should be further considered while designing a pruning criterion is presented, and then an effective definition of diversity is proposed. The experimental results have validated that the given pruning criterion could single out the subset of classifiers that show better performance in the process of hill-climbing search, compared with other definitions of diversity and other criteria.