二维色谱柱效的支持向量回归预测  

Column Efficiency Prediction of Two-Dimensional Chromatography Based on Support Vector Regression

作　　者：蔡从中[1] 温玉锋[1] 裴军芳[1] 朱星键[1] 

机构地区：[1]重庆大学应用物理系,重庆400044

出　　处：《分析化学》2009年第6期835-839,共5页Chinese Journal of Analytical Chemistry

基　　金：教育部新世纪优秀人才支持计划(No.NCET-07-0903);教育部留学回国人员科研启动基金(No.教外司留[2008]101-1);重庆市自然科学基金(No.CSTC;2006BB5240)资助项目

摘　　要：以有效塔板数作为二维色谱的柱效指标,根据二维色谱在不同影响因素(包括预柱柱温、主柱柱温、柱间压差和主柱间的放空量)下的有效塔板数实测数据集,应用基于粒子群算法(PSO)寻优的支持向量回归(SVR)方法,建立了二维色谱柱效的SVR预测模型,并与BP神经网络(BPNN)模型进行了比较。结果表明:基于相同的训练样本和检验样本,二维色谱的SVR模型的平均绝对百分误差(MAPE,13.3%)比其BPNN模型的MAPE小4%;增加训练样本数有助于提高支持向量回归(SVR)模型的泛化性能;基于留一交叉验证法(LOOCV)的SVR模型预测的平均绝对误差(MAE,196.79 m-1)和MAPE(1.6%)均为最小,明显优于BPNN模型(2397.98 m-1,17.3%)或SVR模型(1849.95 m-1,13.3%)的预测效果。因此,SVR是一种预测二维色谱柱效的有效方法。The effective plate number represents the column efficiency of two-dimensional chromatography. Based on the experimental dataset, the support vector regression （SVR） approach combined with particle swarm optimization（PSO） for its parameter optimization was proposed to establish a SVR model for estimating the column efficiency of two dimensional chromatography under different factors, which consisted of tempera- ture of pre-column, temperature of main column, pressure difference between the columns and the vent rate. The prediction results demonstrated that the mean absolute percentage error（ MAPE, 13.3% ） of SVR model was less 4% than that of BP neural network （BPNN） by applying identical training and test samples. For SVR, it was revealed that the estimated error, such as mean absolute error（MAE） and MAPE, both could be efficiently reduced by increasing the number of training samples. The smallest MAE（196.79m-1） and MAPE （1.6%） provided by leave-one-out cross validation （LOOCV） test of SVR were significantly smaller than those achieved by either BPNN（2397.98 m-1, 17.3% ） or SVR（1849.95m-1, 13.3% ）. The results suggest that SVR is an effective and powerful technique for the column efficiency estimation of two-dimensional chromatography.

关 键 词：二维色谱 柱效 支持向量回归 粒子群算法 回归分析 预测 

分 类 号：O657.7[理学—分析化学]