结肠癌基因表达谱的特征选取研究  

Feature selection of gene expression profiles of colon cancer

作　　者：潘冬寅[1] 朱发[1] 徐昇[1] 业宁[1] 

机构地区：[1]南京林业大学信息科学与技术学院,江苏南京210037

出　　处：《山东大学学报（工学版）》2012年第2期23-29,共7页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金资助项目(30671639);江苏省自然科学基金资助项目(BK2009393);江苏省青蓝工程学术带头人资助项目;江苏省科技创新工程资助项目(CXLX11-0525)

摘　　要：为了找到与结肠癌相关的基因,提高结肠癌样本的识别率,提出了基于Chernoff距离的浮动顺序搜索算法(sequential floating search method,SFSM)。通过对结肠癌基因表达谱数据集的分析,对每个基因进行评价和筛选;对筛选后的基因子集利用SFSM算法进行搜索,并以Chernoff距离作为其评估函数,生成若干候选特征基因子集;利用支持向量机(support vector machine,SVM)、K-近邻(K-nearest neighbor,KNN)和径向基(radical basis function,RBF)神经网络分类器来检验候选特征基因子集的分类效果。实验结果表明,利用SFSM及评估函数Chernoff距离发现在参数β=0.25时能找到最佳的特征基因组合,该组合能以很高的正确率识别结肠癌样本。In order to improve the recognition rate of colon cancer sample by selecting the related genes, sequential floating search method（SFSM） basing on Chernoff distance was proposed. Every gene was evaluated and selected by analyzing the data set of the colon cancer gene expression profiles. Some candidate feature gene subsets were obtained by searching the selected gene subset with the method of SFSM whose evaluation function was Chernoff distance. Three different classifies, support vector machines, K-nearest neighbors, and RBF neural networks, were used to validate the classified efficiency. The experimental results showed that when β = 0. 25, the feature gene combination obtained by SF- SM with Chernoff distance as its evaluation function was optimal, and colon cancer sample could be recognized best.

关 键 词：特征选择 Chernoff距离 浮动顺序搜索 支持向量机 K-近邻 径向基神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]