协同区间偏最小二乘 siPLS算法是 N rgaard等对其提出
的 iPLS方法的改进, 其基本算法步骤如下:(1)对原始谱图进行
预处理;(2)在全谱范围内建立全局偏最小二乘模型, 即上节的
模型;(3)在整个光谱区间采用 iPLS建立多个等窗口宽度的子
区间, 假设为 n个;(4)在每个子区间上建立偏最小二乘法模型,
即可得到 n个局部模型;(5)以交叉验证时的均方根误差 RMSE
值为各模型的精度衡量标准, 比较全光谱模型和各局部模型的
精度;(6)组合精度最高的局部子区间为新的回归区间;再在新
的回归区间建立 siPLS模型(Cooperative interval partial least squares (siPLS) algorithm is proposed by N rgaard et al.

The improvements of the iPLS method are as follows: (1) the original spectrogram is performed.

Preprocessing; (2) establishing a global partial least squares model in the full spectrum, that is, the upper part of the model.

Model; (3) use iPLS to establish multiple equal window widths in the whole spectral interval.

The interval is assumed to be n; (4) a partial least squares model is established on each sub interval.

N local models can be obtained; (5) the root mean square error of cross validation is RMSE

The values are the accuracy criteria of each model, and the comparison between the full spectral model and the local models is made.

(6) the local interval with the highest combined precision is a new regression interval; and then new.

The establishment of siPLS model in the regression interval)