Combining ensemble methods of Bagging, Subagging and Random Subspace for phoneme recognition
Abir Bousmina, Chiraz Jlassi, Najet Arous · 2016
A ‘weak classifier’ is a classifier that performed badly for many raisons. In general, bad performance can be caused by the highly dimensionality of the data and also the instability of the classifier. Ensemble methods has been developed in order to overcome this problems. The most popular are bagging and Random Subspace Methods (RSM). We propose to use a combination of concepts used in Bagging and Random Subspaces Methods (RSM) to make five approaches. We test this idea with Support Vector Machines (SVM). Experimental performance indicates that all proposed approaches are effective on solving a phoneme recognition problem and improves the performance of a single SVM.