Anzeige
Erschienen in:
2010 | OriginalPaper | Buchkapitel
1-vs-Others Rough Decision Forest
verfasst von : Jinmao Wei, Shuqin Wang, Guoying Wang
Erschienen in: Rough Set and Knowledge Technology
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Bootstrap, boosting and subspace are popular techniques for inducing decision forests. In all the techniques, each single decision tree is induced in the same way as that for inducing a decision tree on the whole data, in which all possible classes are dealt with together. In such induced trees, some minority classes may be covered up by others when some branches grow or are pruned. For a multi-class problem, this paper proposes to induce individually the 1-vs-others rough decision trees for all classes, and finally construct a rough decision forest, intending to reduce the possible side effects of imbalanced class distribution. Since all training samples are reused to construct the rough decision trees for all classes, the method also tends to have the merits of bootstrap, boosting and subspace. Experimental results and comparisons on some hard gene expression data show the attractiveness of the method.