A selective ensemble classification method on microarray data

Tao Chen

Abstract

For the characteristics of small samples and high dimension of microarray data, this paper proposes a selective ensemble method teaching-learning-based optimization based to classify microarray data. Firstly, in order to remove irrelevant genes with classification task, reliefF algorithm is used to reduce original gene set, and then a new training set is produced from orginal training set according to top-ranked genes obtained. Secondly, multiple bootstrap training subsets are produced based on bagging algorithm on above obtained training set to train base classifiers. Finally, multiple base classifiers are selected by using teaching-learning-based optimization to build an ensemble classifier. Experimental results on eight microarray datasets show our proposed method is effective and efficient for microarray data classification.