多模型共识偏最小二乘法用于近红外光谱定量分析

将多模型共识偏最小二乘法用于近红外光谱定量分析。利用随机抽取的训练子集建立一系列偏最小二乘模型,选取其中性能较好的部分模型作为成员模型,用这些成员模型来预测未知样品。将该方法用于一组生物样本的近红外光谱与样品中人血清白蛋白、γ-球蛋白以及葡萄糖含量之间的建模研究,并与单模型偏最小二乘法了进行比较。结果 PLS对独立测试集中三种组分进行50次重复预测的平均RMSEP分别为0.1066,0.0853和0.1338,RMSEP的标准偏差分别为0.0174,0.0144和0.0416;而本方法重复预测的平均RMSEP分别为0.0715,0.0750和0.0781,RMSEP的标准偏差分别为0.0033,0.2729×10-4和0.0025。

Partial least squares regression method based on consensus modeling for quantitative analysis of near infrared spectra

Partial least squares regression method based on consensus modeling was used for quantitative analysis of near infrared spectra(NIRS).A series of PLS models were built on training subsets which were constructed by random sampling from the training set;the models with high performance were selected as member models,and were used for prediction.The proposed method was used for modeling on NIRS data of a set of biological samples,in which the concentration of human serum albumin(HSA),γ-globulin and glucose were analyzed.Meanwhile,the proposed method was compared with the single-model PLS.As results,mean RMSEP on 50 repeat prediction for HSA,γ-globulin and glucose for the independent test set were 0.1066,0.0853 and 0.1338 by the single-model PLS,the standard deviations of the RMSEPs were 0.0174,0.0144 and 0.0416,respectively;While the mean RMSEP and the corresponding standard deviations by the method proposed in this paper were 0.0715,0.0750,0.0781 and 0.0033,0.2729×10-4,0.0025 respectively.