Elimination of the uninformative calibration sample subset in the modifiedUVE (Uninformative Variable Elimination)-PLS (Partial Least Squares) method

In order to increase the predictive ability of the PLS (Partial Least Squar es) model, we have developed a new algorithm, by which uninformative sample s which cannot contribute to the model very much are eliminated from a cali bration data set. In the proposed algorithm, uninformative wavelength (or i ndependent) variables are eliminated at the first stage by using the modifi ed UVE (Uninformative Variable Elimination)-PLS method that we reported pre viously. Then, if the prediction error of the ith (1 less than or equal to/ less than or equal ton) sample is larger than 3 sigma, the corresponding sa mple is eliminated as uninformative, where n is the total number of calibra tion samples and sigma is the standard deviation calculated from the other n-1 samples. Calculation of sigma by the leave-one-out manner enhances the ability to identify the uninformative samples. The final PLS model is const ructed precisely because both uninformative wavelength variables and uninfo rmative samples are eliminated. In order to demonstrate the usefulness of t he algorithm, we have applied it to two kinds of mid-infrared spectral data sets.