摘要
随着计算机科学和生物医学的发展,基因表达谱数据能够以高维数据的形式导出,这为应用数据挖掘算法对其分析处理提供了可能.基因表达谱数据存在高维度和高冗余特性,实际应用中常需要进行维度约简.基于非负矩阵分解的方法常被选择为维度约简的手段,但由于传统方法未进行针对性处理,在基因数据集上的表现不佳.针对此类数据的特点,本文提出了一种基于非负矩阵分解的改进算法,结合图正则化处理和稀疏化理论,进一步加入了去噪处理,对处理过度冗余的高维基因表达谱数据特别有效.实验表明,算法在肿瘤基因数据集上的表现整体优于传统基于非负矩阵分解的算法.
With the development of computer science and biomedicine,gene expression profile data can be exported in the form of high-dimensional data,which bring possibility to use data mining algorithm to analysis it.Gene expression profile data is normally high dimension and high redundancy,so effective dimension reduction for it in practical application is desirable.Methods based on non-negative matrix factorization（NMF）are frequently used for dimension reduction,but traditional NMF methods perform relatively poor.Therefore,on the basis of the characteristics of gene expression profile data,this paper proposes an improved NMF based algorithm,which combine NMF with graph regularization,sparseness constraints and noise processing to deal with high-dimensional gene data.Experiments results demonstrate that the proposed algorithm performs better than the compared methods on gene expression profile datasets.
作者
张旭
陈志奎
高静
ZHANG Xu;CHEN Zhi-kui;GAO Jing(School of Software Technology,Dalian University of Technology,Dalian 116620,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第11期2480-2484,共5页
Journal of Chinese Computer Systems
基金
国家重点研发计划项目(2016YFD0800300)资助
关键词
非负矩阵分解
维度约简
基因表达谱
稀疏约束
图正则
去噪处理
non-negative matrix factorization
dimension reduction
gene expression profile
graph regularization
sparseness constraints
noise processing
