论文部分内容阅读
为了明确加权函数对变量筛选结果的影响,本文以乳块消糖衣片包衣厚度为载体,用CARS方法筛选其近红外光谱特征变量,并考察加权函数对变量筛选结果的影响。结果表明,借助CARS法可以筛选出CHn在近红外区的特征性吸收变量,模型的预测性能较全谱片最小二乘(PLS)模型有一定的提高。将CARS中原加权函数(i.e.回归系数)替换为VIP后,所选变量的解释性得到提高,但模型的预测性能略有下降。即变量的解释性与模型的预测性能之间没有必然的联系。为了确保结果的可靠性,用上述方法对玉米蛋白特征变量进行筛选,可以得到类似的结论。
In order to clarify the influence of weighting function on the results of variable screening, we used CARS method to screen the characteristic variables of near infrared spectroscopy (FTIR), and investigated the influence of weighted function on the results of variable screening. The results show that the CARs method can be used to screen out the characteristic absorption variables of CHn in the near infrared region. The predictive performance of the model is better than the full-spectrum least squares (PLS) model. After replacing the CARS Central Plains weighting function (the regression coefficient) with VIP, the explanatory variables of the selected variables are improved, but the predictive performance of the model is slightly reduced. There is no necessary relation between the explanatory power of the variables and the predictive performance of the model. To ensure the reliability of the results, a similar conclusion can be drawn by screening the zymogram variables using the above method.