【摘要】 对分光光度同时测定润滑油中的Ca,Ba,原油中的Fe,Ni,V,润滑油中的Fe,Cu,Zn和铝合金中的Fe,Mn,Cu,Zn的光谱数据分别采用偏最小二乘(PLS)和ε-支持向量机(ε-SVM)两种方法进行解析,结果表明PLS和ε-SVM都能利用校正样建立有效的校正模型对合成样进行合理预测,但从预测结果的绝对误差和平均相对误差的比较看,ε-SVM的预测准确率要比PLS方法高,表明ε-SVM在紫外光谱数据解析方面有着比PLS更好的回归能力,适合用来处理多元校正问题.更多还原

【Abstract】 Support Vector Machine(SVM),a machine learning technology based on statistical learning theory(SLT) and structure risk minimization,has excellent regression ability,for it can find global solution and work with high dimensional input vectors.However,Support Vector Machine(SVM) is seldom reported compared with the widely known Partial Least Squares(PLS) in chemometrics.This paper compares the use and the performance of PLS and SVM for four spectral regression applications.Furthermore,linear interpolation is used to increase the number of training samples,which can compensate lacking of information in some extensions.The results show that the performance of SVM is better than that of PLS for both raw and interpolated spectral data,and that the prediction accuracy of interpolated spectral data is better than that of the raw spectral data.更多还原