基于小波高频系数基因芯片数据的特征提取  

Wavelet high-frequency coefficients for feature extraction of gene microarray data

作　　者：刘玉杰[1] 刘毅慧[1] 

机构地区：[1]山东轻工业学院信息科学与技术学院智能信息处理研究所

出　　处：《生物信息学》2011年第4期339-343,共5页Chinese Journal of Bioinformatics

基　　金：山东省自然科学基金(Y2008G30)

摘　　要：结合小波分析理论与支持向量机理论,构造分类器模型,将前列腺癌基因芯片数据分成癌症和正常两种。本文着重研究小波高频系数基因芯片数据的特征提取,并通过实验对比小波高频系数和低频系数特征提取对分类器性能的影响。其中haar小波3层分解提取高频系数,送入分类器分类后,得到的正确分类率为93.31%。db1小波4层分解提取低频系数,送入分类器分类后,得到的正确分类率为93.53%。小波低频系数特征提取分类效果总体上好于高频系数,分类器性能稳定。In the paper,we use the wavelet analysis theory and the support vector machine theory to build a model which can classify the prostate cancer microarray data into cancer and normal classes.We mainly research the wavelet high-frequency coefficients for feature extraction of prostate cancer gene microarray data in contrast to the low coefficients.We extract haar wavelet high-frequency coefficients at level 3 and feed the high-frequency coefficients to the classification.The correct classification rate is 93.31%.We extract db1 wavelet low-frequency coefficients at level 4 and feed the low-frequency coefficients to the classification.The correct classification rate is 93.53%.The wavelet low-frequency coefficients for feature extraction are better than high-frequency coefficients.The classification modle is very stable.

关 键 词：小波分析 支持向量机 前列腺癌基因芯片数据 低频系数 高频系数 

分 类 号：Q617[生物学—生物物理学] TP391[自动化与计算机技术—计算机应用技术]