中文摘要：

采用CARS（Competitive adaptive reweighted sampling）变量筛选方法建模,显著提高了液态奶中蛋白质与脂肪近红外模型的预测精度。用蒙特卡罗采样（Monte-Carlo sampling）方法先剔除奇异样本,再对光谱进行中心化与Karl Norris滤波降噪处理,通过CARS方法筛选出与样本性质密切相关的变量,建立预测蛋白质与脂肪含量的偏最小二乘法（PLS）校正模型,并与未选变量的PLS模型进行比较。以定标集相关系数（r2）及交互验证均方残差（RMSECV）和预测误差均方根（RMSEP）作为判定依据,确定了蛋白质与脂肪的最佳建模条件。蛋白质与脂肪校正模型的相关系数分别为0.975 0、0.995 1,RMSECV分别为0.194 8、0.136 3,RMSEP分别为0.113 3、0.140 1,预测结果优于未选变量的PLS模型及其他选变量方法,有效简化了模型,适于液态奶中脂肪和蛋白质的快速、无损检测。

英文摘要：

Near-infrared（NIR） spectroscopy has become an extensively used analytical technique in many industrial applications because of its rapidness and the fact that it is non-destructive to the samples.In this paper,competitive adaptive reweighted sampling method was employed to improve the prediction accuracy of the NIR model of protein and fat in liquid milk.Monte-Carlo sampling method was employed to detect the outlier before preprocessing the spectroscopy by means of centering and Karl Norris derivative method.The key variables which closely related to the nature of the samples were selected by competitive adaptive reweighted sampling method.The partial least squares（PLS） calibration models were established in the optimal conditions to predict the content of protein and fat,and compared with the results not using CARS.The results showed that better prediction was obtained by CARS when compared to full spectrum PLS modeling,Monte-Carlo uninformative variable elimination（MC-UVE） and moving window partial least squares regression（MWPLSR）.Correlation coefficient（r2）,root-mean-square error of cross-validation（RMSECV） and root-mean-square error of prediction（RMSEP） were used to evaluate the quality of the model.The best models showed satisfactory predictions as measured by r^2,RMSECV and RMSEP values： protein,0.975 0,0.194 8 and 0.113 3;fat,0.995 1,0.136 3 and 0.140 1,respectively.The result showed that using variable selection method such as CARS could effectively simplify the model and solve practical problems.The proposed method was suitable for the fast and reliable determination of protein and fat in liquid milk.