The research of multi-class problem based gene expression profile data

针对基因表达谱微阵列的数据多分类问题,给出一种在多病类情况下的基于信噪比和相关性的特征基因选择方法.该方法一次性考虑基因区分所有病类的能力,尽量避免基因的冗余性;其次利用支持向量机,构建了基因表达谱微阵列数据的多分类器;最后通过实验表明了本方法的有效性.Aiming at the multi-class problem of gene expression profile data,this paper proposes a gene selection method based on S2N and correlation for multiple diseases.This method takes the classification abilities of genes to separate all the diseases into consideration at a time and tries to avoid redundancy in selected genes.Secondly,we construct multi-classifier of gene expression profile data using SVM.Finally,we do experiment by this method,the result of which shows great effectiveness of the method.国家自然科学基金资助项目(60704042