Multi-way partial least squares modeling for three-dimensional fluorescence spectrometry

As extension of partial least square(PLS) to high order,multi-way partial least squares(N-PLS) incorporate the dependent variables in the decomposition of the independent variables.The data of fluorescence spectrometry for samples is arranged in a three-way array,where the first refers to samples,the second to excitation wavelengths,and the third to emission wavelengths.And the N-PLS is used to develop calibration model on the basis of the constructed three-way data.Experimental results for detecting TOC and COD of the dissolved organ matters show N-PLS has an advantage over PARAFAC and conventional PLS in regress modeling.