在润滑油酸值进行红外光谱法测定中，采用最小二乘支持向量机（LS-SVM）建立了酸值的定量预测模型。用Kennard-Stone方法将30个样本划分为训练集（24个样本）和验证集（6个样本），进行定量预测，并与偏最小二乘法和径向基函数神经网络所建模型的预测进行比较。结果表明：LS-SVM所建模型的预测标准偏差（SEP）最小，仅为0.002；预测值的相对误差为1.3%~5.3%。由此认为LS-SVM所建模型的训练和预测结果均优于其余两种方法所建模型。对5个未知样品的分析结果表明：LS-SVM模型的预测值与化学法实测值之间的相对误差（1.2%~3.1%）也较少。

In the infrared spectrometric determination of acid value of lubricating oil, quantitative prediction model for the acid value was established by the least square vector machine (LS-SVM) method. By applying the Kennard-Stone method, 30 samples were divided into 2 groups, the training group (with 24 samples) and the testing group (with 6 samples), and quantitative prediction was performed. As compared with the models developed by methods of PLS and RBF-NH, the model of LS-SVM gave the smallest value of SEP (0.002), and values of relative error of the predicted values were found in the range of 1.3%-5.3%. It was shown that the results of training and prediction obtained by the model of LS-SVM were better than those obtained by the other 2 models. In the analysis of 5 unknown samples, small relative errors (ranged from 1.2% to 3.1%) were obtained by the model of LS-SVM between the predicted values and the true values.