为了对基因表达数据矩阵中的肿瘤基因与正常基因进行判别分类,文章提出了基于支持向量机(Supporting Vector Machine,SVM)的肿瘤基因识别方法.在对基因进行特征选择的基础上,对只具有最优特征的基因样本再利用SVM分类思想进行判别,最后通过与其他方法所得结果进行对比可知,该方案在不降低分类准确度的同时,能有效地避免特征空间维数远大于样本空间维数所造成的“过学习”问题,而且避免了大的时空开销,具有很强的实用性.%In order to classify the tumor and the normal samples in the gene expression data, an identification method based on Support Vector Machine(SVM) is proposed in this paper. First select the best "feature genes", and then identify the type of the gene sample with the optimal feature genes using SVM classification. The results compared with other methods showed the proposed method can effectively avoid the "over fit problem" that occurred when the dimension of feature space is much larger than that of the sample space. Therefore, the identification scheme is of practical use.
展开▼