Hu, Hong and Li, Jiuyong and Wang, Hua and Daggard, Grant and Shi, Mingren (2006) A maximally diversified multiple decision tree algorithm for microarray data classification. In: Workshop on Intelligent Systems for Bioinformatics, 4 Dec 2006, Hobart, Australia.
|HTML Citation||EndNote||Dublin Core||Reference Manager|
Full text available as:
|PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader|
We investigate the idea of using diversified multiple trees for Microarray data classification. We propose an algorithm of Maximally Diversified Multiple Trees (MDMT), which makes use of a set of unique trees in the decision committee. We compare MDMT with some well-known ensemble methods, namely AdaBoost, Bagging, and Random Forests. We also compare MDMT with a diversified decision tree algorithm, Cascading and Sharing trees (CS4), which forms the decision committee by using a set of trees with distinct roots. Based on seven Microarray data sets, both MDMT and CS4 are more accurate on average than AdaBoost, Bagging, and Random Forests. Based on a sign test of 95% confidence, both MDMT and CS4 perform better than majority traditional ensemble methods tested. We discuss differences between MDMT and CS4.
Archive Staff Only: edit this record