安徽大学学报（自然科学版）2011，Vol.35Issue(4)：86-91,6.
基于邻接谱主分量分析的肿瘤分类方法
Tumor classification based on principal components analysis of adjacency spectrum
摘要
Abstract
Based on spectral graph theory, a gene expression data classification research was proposed. The features that could reflect the graph structures information were introduced into gene expression data classification, so that the discrete tumor sample dots with high dimension could be transformed into the graphs with rich structural information. In this paper, the authors constructed the Gauss adjacency matrix on gene expression samples, and by means of singular value decomposition, found out the main component that could to the utmost distinguish tumor samples and normal samples using feature scoring criteria. Then they put this component as sample characteristics into the KNN classifier. The experiment on leukemia gene expression data (AML and ALL) and colon cancer gene expression data showed the feasibility and effectiveness of this method.关键词
肿瘤分类/主分量分析/邻接矩阵/特征记分准则Key words
tumor classification/ principal components analysis/ adjacent spectral/ feature scoring criteria分类
信息技术与安全科学引用本文复制引用
陈乐,王年,苏亮亮,王蕊平..基于邻接谱主分量分析的肿瘤分类方法[J].安徽大学学报（自然科学版）,2011,35(4):86-91,6.基金项目
国家自然科学基金资助项目(60772121) （60772121）
