A GA-based feature selection and ensemble learning for high-dimensional datasets
Pei-Yong Xia, Xiangqian Ding, Bai-Ning Jiang · 2009
When dealing with high-dimensional datasets with fewer samples, feature selection and ensemble learning are two effective strategies. In this paper, we focus our attention on genetic based feature selection for ensemble learning. We use an improved GA algorithm (IGA) to reduce the dimensionality of the feature space, and then evaluate using Bagging and Ada-Boost constructed by the reduced features. Experimental results on several UCI datasets demonstrate that the improved GA-based feature selection algorithm (IGAFS) is often able to obtain a better feature subset when compared with the standard GA-based feature selection algorithm (SGAFS). Our experiments also indicate that ensemble learning using IGAFS is more accuracy than employing SGAFS and the whole feature space in general conditions.