摘要
针对基因表达谱高维、小样本、高冗余和高噪声等特点,提出了一种特征基因混合选择方法.采用Relief F方法对原始基因进行排序,过滤无效基因,获得初选基因子集,给出了基于差分进化算法优化的核模糊粗糙集模型,进行了特征基因终选.仿真实验结果表明:所提算法比Relief F、Kruskal Wallis、Gini Index等算法在分类精度和基因数量等方面有明显优势.
A hybrid feature gene selection algorithm based on an improved kernelized fuzzy rough sets aiming at the characteristics of high dimensions,small samples,high noise and high redundancy of gene expression profiles is proposed.The top-ranked genes based on Relief F algorithm are selected to construct a primary gene subset in order to remove the invalid genes. An improved kernelized fuzzy rough sets model based on the differential evolution algorithm is proposed to achieve the selection of feature genes. Simulation results show that the proposed algorithm has obvious advantages comparison with Relief F,Kruskal Wallis and Gini index algorithm.
作者
陈涛
Chen Tao(School of Mathematics and Computer Science,Shanxi University of Technology,Hanzhong,72300)
出处
《中南民族大学学报（自然科学版）》
CAS
2018年第2期121-127,共7页
Journal of South-Central University for Nationalities：Natural Science Edition
基金
国家自然科学基金资助项目(11502132)
陕西省教育厅科研资助项目(16JK1149)
陕西理工大学科研资助项目(SLGQD2017-07)
