【摘要】 为了对癌症等疾病分型、诊断及进行病理学研究,利用基因微阵列数据识别疾病相关基因.考虑到了基因微阵列数据是典型的矛盾决策系统,在证明矛盾系统在近似分布集上是协调的这一事实的基础上提出了一套近似分布约简理论,讨论了不同近似分布集上约简之间的关系,提出了基于近似约简的基因选择方法.使用两组真实的基因表达数据对所提出的方法进行了验证.实验结果表明,该方法能在保持分类能力的情况下降低特征基因集的相关性,从而显著地减少特征基因的数量.更多还原

【Abstract】 To identify disease related genes from gene expression profiles(DNA microarray) has a very important practical significance for disease,such as cancer,subtype discovery,diagnosis and pathology study.Gene selection is a critical preprocessing technique for the DNA microarray data analysis.Gene sets of interest typically selected by usual ranking methods from DNA microarray data will contain many highly correlated genes and remain high dimension.Thinking of that DNA microarray data sets are typical inconsistent decision system,with introduction of the concept of that a inconsistent decision system is consistent according to its approximate distribution,a set of notions of approximate distribution reduct are proposed.After discussed the relations between the lower and upper approximation reducts of a inconsistent decision system,a gene selection method based on approximate distribute reduct is obtained.The experimental results on two publicly available DNA microarray datasets,lung cancer and NCI60(9-tumors),show that the proposed method got an equivalent classification effect with significantly reduced number of selected gene.更多还原