Research Article
A New Ensemble Pruning Method Based on Margin and Diversity
@INPROCEEDINGS{10.1007/978-3-030-89814-4_50, author={Zixiong Shen and Xingcheng Liu}, title={A New Ensemble Pruning Method Based on Margin and Diversity}, proceedings={Mobile Multimedia Communications. 14th EAI International Conference, Mobimedia 2021, Virtual Event, July 23-25, 2021, Proceedings}, proceedings_a={MOBIMEDIA}, year={2021}, month={11}, keywords={Ensemble learning Classification Multiple classifier systems}, doi={10.1007/978-3-030-89814-4_50} }- Zixiong Shen
Xingcheng Liu
Year: 2021
A New Ensemble Pruning Method Based on Margin and Diversity
MOBIMEDIA
Springer
DOI: 10.1007/978-3-030-89814-4_50
Abstract
Classification is one of the main tasks of machine learning, and ensemble learning has become a successful paradigm in the data classification field. This work aims to present a new method for pruning an ensemble classification model based on margin theory and ensemble diversity. Firstly, a new unsupervised form of instances margin metric is proposed, which does not need to consider the true class labels of the instances. This mechanism can improve the robustness of the algorithm against mislabeled noise instances. Then, the Jensen-Shannon (J-S) divergence between the classifiers is calculated based on the probability distribution of the class labels. Finally, all base classifiers are ordered with respect to a new criterion which combines the obtained margin values and the J-S divergence of base classifiers. Experiments show that the proposed method has a stable improvement on a significant proportion of benchmark datasets over existing ensemble pruning methods.
