A new approach to improving ICA-based models for the classification of microarray data

Abstract

Inspired by the idea of ensemble feature selection, we design an ICA based ensemble learning system to fully utilize the difference among different IC sets. Firstly, some IC sets are generated by different ICA transformations. A multi-objective genetic algorithm (MOGA) is then designed to select different biologically significant IC subsets from these IC sets, which are applied to build base classifiers. In addition, a global-recording technique is designed to record the best IC subsets of each IC set discovered by the MOGA into a global-recording list. When MOGA stops, all individuals in the list are deployed to train base classifiers. The base classifiers generated by these schemes are fused by the majority vote rule. Three microarray datasets are used to test the ensemble systems, and the corresponding results demonstrate that two ensemble schemes can improve the performance of the ICA based classification model. © 2009 Springer Berlin Heidelberg.