基于非线性支持向量机的原核生物基因识别  

Prokaryotes gene identification based on nonlinear SVM

作　　者：张继宏[1] 李小霞[1] 孙波[1] 

机构地区：[1]西南科技大学信息工程学院,四川绵阳621010

出　　处：《计算机应用》2009年第10期2748-2750,共3页journal of Computer Applications

摘　　要：应用非线性最小二乘支持向量机对原核生物进行基因识别，通过寻找序列开放阅读框（ORF），并与可靠基因位点文件进行比较产生训练样本集，然后提取样本GC含量和Z曲线特征，并利用T检验方法检验各特征值所包含的信息量，设计出了非线性最小二乘支持向量机分类器识别基因。结果表明非线性最小二乘支持向量机的识别率比Fisher判别和线性支持向量机在不同的特征组合下分别提高了7．09％-29．97％和10．97％-25．45％，并且在特征值信息量较小的情况下非线性最小二乘支持向量机更能表现其优越性。This paper presented a nonlinear least squares support vector machine method to identify the prokaryotes gene. This method generated training sample sets by searching sequence Open Reading Frames （ORF） and comparing ORF sets with reliable gene location document, extracted two sample features of GC content and Z curve, examined information content of these features by T-test, and designed nonlinear least squares support vector machine classifier to recognize gene. The results show that the recognition rates of nonlinear least squares support vector machine are 7.09% - 29.97% and 10.97% -25.45% higher than Fisher and linear support vector machine respectively under different feature combinations, and the nonlinear support vector machine method performs better when the feature information content is less.

关 键 词：基因识别 非线性最小二乘支持向量机 原核生物 GC含量 Z曲线 T检验 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]