主成分分析-支持向量机在卷烟主流烟气分析中的应用  

Establishment and application of principal component analysis-support vector regression model in cigarette mainstream smoke analysis

作　　者：章平泉[1] 龚珍林[1] 杜秀敏[1] 俞京[1] 金岚峰[1] 

机构地区：[1]江苏中烟工业有限责任公司淮阴卷烟厂品质管理处,淮安市一品梅路32号223002

出　　处：《中国烟草学报》2010年第6期21-24,共4页Acta Tabacaria Sinica

摘　　要：采用主成分分析进行样本集特征的提取,结合支持向量机建立回归模型,并对成品卷烟主流烟气中的总粒相物、焦油量和烟气烟碱含量进行定量预测。结果表明:总粒相物、焦油和烟气烟碱的预测均方差分别为0.61,0.47和0.04,与模型相比分别下降了30.73%,26.12%和8.15%,体现了更高的预测准确度。A new quantitative regression model was built in combination with extracts feature from sample cluster using principal component analysis（PCA）.A nonlinear model was built using support vector regression（SVR） to predict total particulate matter,tar and nicotine content in cigarette smoke.Results showed that the RMSEP（root mean square error of prediction） of total particulate matter,tar and nicotine in smoke was 0.61,0.47 and 0.04 respectively,reduced by 30.73%,26.12% and 8.15% compared with SVM method,indicating that PCA-SVM method resulted in high prediction accuracy.

关 键 词：主成分分析 支持向量机 总粒相物 焦油 烟气烟碱 

分 类 号：TS411.2[农业科学—烟草工业]