A Variable Selection Method for Near-infrared Spectroscopy Based on Iterative Shrinkage Window Bootstrapping Soft Shrinkage Algorithm

KeyWord:variable selection  iterative shrinkage window  near-infrared spectroscopy  partial least squares  model calibration

Abstract:

      In this paper£¬an iterative shrinkage window-bootstrapping soft shrinkage£¨ISWBOSS£© algorithm was proposed to address problems of poor prediction performance of the calibration models based on near-infrared£¨NIR£© spectroscopy due to their redundant variables in NIR spectra£®The variables was divided by windows in the method£¬and the windows were randomly selected£¬in which the sub-models were built with the variables£®The soft shrinkage of the variable space was gradually achieved by calculating the normalization of the regression coefficients of the variables in the window£¬and continuing the weighted sampling as weights£®Meanwhile£¬the window size was continuously shrunk during the iterative process to perform an accurate search of the feature variables£®It was validated on a corn dataset£¬and compared with the partial least squares models established by the full-spectrum method£¬genetic algorithm£¬competitive adaptive reweighted sampling£¬and bootstrapping soft shrinkage approach£®The results showed that the new method had significant advantages in terms of both accuracy and stability£®Taking corn protein content prediction as an example£¬the root mean square error of prediction of ISWBOSS was reduced from 0.041 8 to 0.010 3£¬compared with the bootstrapping soft shrinkage approach£®Moreover£¬the new method required fewer iterations and higher operational efficiency to reach the optimal model£¬which was a guideline for improving the performance of NIR spectral calibration models£®