CAJ | 학술논문

针对肿瘤基因表达谱具有高维度、小样本的特点，提出一种面向基因表达数据的混合特征选择方法。首先利用 Relief 算法从原始基因集中获取10％左右与分类相关的基因作为相关基因子集，然后利用 SVM -RFE 算法对相关基因子集再进行选择，获取对分类影响较大的排序基因子集，最后利用 Wrapper 方法对排序基因子集进行选择，获得最佳基因子集。以支持向量机为分类器，对急性白血病、结肠癌肿瘤表达谱数据进行测试，实验结果表明所提方法时间开销较小，且取得较高的分类精度。
침대종류기인표체보구유고유도、소양본적특점，제출일충면향기인표체수거적혼합특정선택방법。수선이용 Relief 산법종원시기인집중획취10％좌우여분류상관적기인작위상관기인자집，연후이용 SVM -RFE 산법대상관기인자집재진행선택，획취대분류영향교대적배서기인자집，최후이용 Wrapper 방법대배서기인자집진행선택，획득최가기인자집。이지지향량궤위분류기，대급성백혈병、결장암종류표체보수거진행측시，실험결과표명소제방법시간개소교소，차취득교고적분류정도。
A bstract:According to the characteristics of the tumor gene expression profiles which includes high dimensional-ity and small samples,this paper proposes a hybrid features selection method for gene expression data.Firstly, relief algorithm to get about 10% of the original gene set which is related to classification as the relevant gene subset,and then is used SVM-RFE algorithm to obtain the sorted gene subset that has greater impact on the classification from the relevant gene subset.Finally,is used the Wrapper method to select the optimal gene sub-set from the sorted gene subset.Using support vector machine as classifier,is used acute leukemia and colon can-cer microarray datasets for testing,experimental results show that the proposed method costs less time,and achieves higher classification accuracy.