一种改进的基于先验信息和微粒群算法的基因选择方法  

An improved gene selected method based on priori information and PSO

作　　者：凌青华[1,2] 孙伟[2] 韩飞[2] 

机构地区：[1]江苏科技大学计算机科学与工程学院,镇江212003 [2]江苏大学计算机科学与通信工程学院,镇江212013

出　　处：《江苏科技大学学报（自然科学版）》2017年第6期774-780,共7页Journal of Jiangsu University of Science and Technology:Natural Science Edition

基　　金：国家重点研发计划项目(2017YFC086600);江苏省六大人才高峰人才计划项目(2015-DZXX-24);江苏省"333高层次人才培养工程"项目((2016)III-0845)

摘　　要：针对KMeans-GCSI-MBPSO-ELM方法中冗余基因去除时会出现的"误删除"现象,提出一种改进的基于先验信息和微粒群算法(particle swarm optimization,PSO)的基因选择方法(I-KGME).在充分考虑基因类别灵敏度(gene-to-class sensitivity,GCS)信息的基础上,利用二进制PSO耦合GCS信息和K-均值聚类算法进行基因选择,并通过二进制PSO算法实现优化.该类方法能够获取低冗余、高预测性的基因子集,并在多个基因表达谱数据上获得了优于经典基因选择方法和KMeansGCSI-MBPSO-ELM方法的性能.Considering the ＂deleting by mistake＂Occurring in the procedure to remove redundant genes in the KMeans-GCSI-MBPSO-ELM method,we proposed an improved algorithm I-KGME based on priori information and PSO（particle swarm optimization）. On the basis of considering sufficiently gene-to-class sensitivity（GCS）information,we combined binary particle swarm optimization（BPSO） to GCS and K-means clustering method to perform gene selection. And in the same time,we realized optimization by BPSO. The new gene selection method could obtain the gene subsets with low redundancy and high prediction ability. The experiment results on some gene express profile data verify that the proposed method obtains better performance than some classic methods and KMeans-GCSI-MBPSO-ELM.

关 键 词：基因选择 基因类别灵敏度信息 K-均值 微粒群优化 基因表达谱数据 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]