Neural network ensemble pruning using sensitivity measure in web applications

@inproceedings{citeulike:2292153, abstract = {Multiple Classifier Systems (MCSs) have been shown theoretically and empirically to outperform a single classifier in many applications. However, many ensemble training algorithms sometimes create a very large MCS which is combined by many individual classifiers. A large MCS not only consumes computational resources but also decreases the effectiveness. One of the solutions is the pruning method. It reduces the number of individual classifiers inside an MCS that maintains the performance well or is just slightly worse than the original one. In this paper, a new pruning method, called NNEPSM, for Neural Network ensemble based on a sensitivity measure is proposed. The classifiers which have less impact to the final output of MCS will be removed. The advantages of this method include efficient performance, low-complexity and independence on training method. NNEPSM has been applied in Web applications and other benchmark dataset. The experimental results showed that our approach performs well using different datasets.}, author = {Chan, Patrick P. and Zeng, Xiaoqin and Tsang, Eric C. and Yeung, Daniel S. and Lee, John W. }, booktitle = {Systems, Man and Cybernetics, 2007. ISIC. IEEE International Conference on}, citeulike-article-id = {2292153}, doi = {10.1109/ICSMC.2007.4413821}, journal = {Systems, Man and Cybernetics, 2007. ISIC. IEEE International Conference on}, keywords = {citation, ensemble-selection}, pages = {3051--3056}, posted-at = {2008-01-26 09:05:14}, priority = {0}, title = {Neural network ensemble pruning using sensitivity measure in web applications}, url = {http://dx.doi.org/10.1109/ICSMC.2007.4413821}, year = {2007} }

TY - CONF ID - citeulike:2292153 L3 - citeulike-article-id:2292153 TI - Neural network ensemble pruning using sensitivity measure in web applications T2 - Systems, Man and Cybernetics, 2007. ISIC. IEEE International Conference on JF - Systems, Man and Cybernetics, 2007. ISIC. IEEE International Conference on SP - 3051 EP - 3056 N2 - Multiple Classifier Systems (MCSs) have been shown theoretically and empirically to outperform a single classifier in many applications. However, many ensemble training algorithms sometimes create a very large MCS which is combined by many individual classifiers. A large MCS not only consumes computational resources but also decreases the effectiveness. One of the solutions is the pruning method. It reduces the number of individual classifiers inside an MCS that maintains the performance well or is just slightly worse than the original one. In this paper, a new pruning method, called NNEPSM, for Neural Network ensemble based on a sensitivity measure is proposed. The classifiers which have less impact to the final output of MCS will be removed. The advantages of this method include efficient performance, low-complexity and independence on training method. NNEPSM has been applied in Web applications and other benchmark dataset. The experimental results showed that our approach performs well using different datasets. KW - citation KW - ensemble-selection AU - Chan, Patrick AU - Zeng, Xiaoqin AU - Tsang, Eric AU - Yeung, Daniel AU - Lee, John PY - 2007/// UR - http://dx.doi.org/10.1109/ICSMC.2007.4413821 ER -