Refining Transfer Set in Calibration Transfer of Near Infrared Spectra by Backward Refinement of Samples
The transfer set in calibration transfer can be further refined by backward refinement (BR) method. The BR includes many loops, in each of which the sample can be removed which reduces the RMSEV best. Then, this procedure can be repeatedly executed many times so that many subsets of samples can be obtained. Finally, the sample subset with a small value of RMSEV can be chosen as the best sample subset. Two batches of datasets including corn and wheat datasets have been applied to testing this method. The results showed that compared with ordinary Kennard–Stone (KS) method, BR can reduce errors largely when it is applied to calibration transfer methods including canonical correlation analysis combined with informative components extraction (CCA-ICE), direct standardization (DS) and piecewise direct standardization (PDS).