Incorporating gene similarity into support vector machine for microarray classification and gene selection

摘要

In this paper, we propose a novel method based on support vector machine (SVM) formicroarray classification and gene (feature) selection. The proposed method, called similarity-based SVM (SSVM), incorporates the prior knowledge of gene similarity into the standard SVMby combining the standard l_2norm and the similarity penalty of all genes. The preliminaryexperiments show that our method performs better than the standard SVM,l_2 — l_2SVM and SVM-RFE, especially when the features are highly similar.