在癌症分类中基于分层抽样的神经网络集成算法  

A neural network ensemble method based on the stratified sampling in tumor classification

作　　者：钟金贝[1] 林亚平[2] 卢新国[2] 

机构地区：[1]湖南大学软件学院,长沙410085 [2]湖南大学计算机与通信学院,长沙410085

出　　处：《微计算机信息》2010年第4期178-180,共3页Control & Automation

摘　　要：在基因表达谱数据的分析中,针对有效合理地选择特征基因集的问题,本文将分层抽样技术引入特征基因选择,提高特征基因集的分类能力。以神经网络作为分量分类器,神经网络集成进行分类预测。并在结肠癌数据集上进行实验,实验结果表明该方法能有效地降低特征基因集选择的复杂性,提高对于未知样本的分类预测效果。With introducing the stratified sampling into the character gene＇s selection for the problem of choosing gene as charater gene group effectively and rationaly in the analysis of the expression profiles, a feature selection method based on the stratified sam- piing was proposed for improving the classification ability of character genc group. And a network ensemble which neural networks was taken as the individual classification is employed to classify the samples. In the end, this method was experimentize in the colon tumor dataset, and the experiments had shown this method can reduce the complexity of the feature selection,and enhance the effec- tiveness for unkown sample＇s classification.

关 键 词：神经网络集成 基因表达谱 偏度 分层抽样 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] Q786[自动化与计算机技术—计算机科学与技术]