A robust ensemble classification method analysis

Abstract

Apart from the dimensionality problem, the uncertainty of Microarray data quality is another major challenge of Microarray classification. Microarray data contain various levels of noise and quite often high levels of noise, and these data lead to unreliable and low accuracy analysis as well as high dimensionality problem. In this paper, we propose a new Microarray data classification method, based on diversified multiple trees. The new method contains features that (1) make most use of the information from the abundant genes in the Microarray data and (2) use a unique diversity measurement in the ensemble decision committee. The experimental results show that the proposed classification method (DMDT) and the well-known method (CS4), which diversifies trees by using distinct tree roots, are more accurate on average than other well-known ensemble methods, including Bagging, Boosting, and Random Forests. The experiments also indicate that using diversity measurement of DMDT improves the classification accuracy of ensemble classification on Microarray data. © 2010 Springer Science+Business Media, LLC.