两种过滤特征基因选择算法的有效性研究

对基因表达谱进行特征基因选择不仅能改善疾病分类方法的效能,而且为寻找与疾病相关的特征基因提供新的途径．通过比较用调整p值的t检验、非参数评分两种特征基因选择算法后和未进行选择时支持向量机(SVM)分类器的分类性能、支持向量(SV)的吻合度、错分样本ID的吻合度和对样本均匀翻倍后的稳定性．结果发现：特征选择后线性、核函数为二阶多项式和径向基的SVM分类性能明显提高;特征选择前后的SV及错分样本ID的吻合度均较高;SVM的稳定性较好．由此得出结论：这两种特征选择算法具有一定的有效性．

Efficiency of Two Filter Feature Gene Selection Algorithms

Application of feature gene selection in gene expression profile not only improve the efficiency of disease pattern classification methods but also provide a new way to find the genes correlation with disease. The performance of Support Vector Machine classifiers, the consistency of support vectors, the consistency of error samples and equably fold train samples between after using adjust p value t-test, nonparametric scoring and using all genes were compared. The results showed that t-test, linear SVM, square polynomial SVM, Radial Basis Function SVM have excellent performance and the stability of SVM is better after feature selection. The consistency of support vectors and the consistency of error samples were high between pre-feature-selection and post-feature-selection. It could be concluded that the two feature gene selection algorithms have strong validity.