摘要
采用偏最小二乘法(PLS)和主成分分析法(PCA)对光谱数据进行主成分提取,并利用遗传算法(GA)对波长进行选择,剔除不相关的变量,达到压缩数据的目的;采用反向传播人工神经网络(BP-ANN)方法,对多组分体系进行定性定量分析.建立了一个PLS-BP-ANN、PCA-BP-ANN、GA-BP-ANN和BP-ANN四种模型的化学计量学多组分分析平台,对谱带混叠严重的5种大气有机毒物——苯、甲苯、甲醇、氯仿和丙酮进行了定性定量测定,比较了4种模型的误差,结果表明,PLS-BP-ANN模型得到的结果最好.
The principal component of spectrum data is extracted through the methods of partial least squares(PLS) and principal component analysis(PCA),and the wavelength is selected based on genetic algorithm(GA) by removing uncorrelated variables for the sake of data compressing.The qualitative and quantitative analysis of multi-component system is performed through BP-ANN method.The chemo metrics multi-component analysis platform composed of PLS-BP-ANN,PCA-BP-ANN,GA-BP-ANN and BP-ANN models performs the qualitative and quantitative analysis of five air toxics(benzene,toluene,methanol,chloroform and acetone) whose absorption bands overlap seriously.The prediction errors of the four models are compared,which shows that the PLS-BP-ANN model works best.
出处
《南通大学学报(自然科学版)》
CAS
2011年第4期26-30,共5页
Journal of Nantong University(Natural Science Edition)
基金
江苏省高校自然科学基础研究项目(08KJB610007)
南通市科学技术局科技计划项目(S2009010)
关键词
多组分体系
偏最小二乘法
主成分分析
人工神经网络
定性定量分析
multi-component system
partal least squares
principal component analysis
artificial neural network
qualitative and quantitative analysis