论文部分内容阅读
提出了一种基于回归系数的变量逐步筛选方法。对光谱中各变量计算其回归系数后,按其绝对值由大到小将相应变量排列,采用PLS交互检验按前向选择法逐步选择最佳变量子集。用该方法对玉米和柴油近红外光谱数据进行分析,对玉米蛋白质、柴油十六烷值和粘度分别选择出了14、12以及30个最佳变量用于建模,所得预测结果均优于全谱变量建模的预测结果。可见本方法是一种有效实用的近红外光谱变量选择方法。
A variable screening method based on regression coefficient is proposed. After calculating the regression coefficients of each variable in the spectrum, the corresponding variables are arranged according to their absolute values, and the PLS interaction test is used to gradually select the best subset of variables according to the forward selection method. Using this method, the data of corn and diesel near infrared spectroscopy were analyzed. 14, 12 and 30 best variables were chosen for the modeling of corn protein, diesel cetane number and viscosity respectively. The results obtained were all better than the whole Prediction of Spectral Variables Modeling. It can be seen that this method is an effective and practical method for the selection of near-infrared spectral variables.