Usefulness of PARAFAC for the Quantification, Identification, and Description of Analytical Data

Numerous research activities generate data that are organized as a three-way tensor, which can be described with PARAFAC. This chapter starts by briefly describing the PARAFAC model and the properties that support its practical utility. Besides, two alternative models are presented, PARAFAC2 and Tucker3, which are useful when the data are inconsistent with the PARAFAC model. The properties are illustrated throughout four cases that consider different analytical scenarios: (1) quantification of a target analyte in presence of interferents (excitation-emission fluorescence spectra); (2) quantification and identification of an analyte for regulated analyses ensuring the probabilities of false positive and false negative, even in the presence of interferents (gas chromatography-mass spectrometry data for pesticides); (3) prediction of a quantitative response with no specific signals (electronic nose data to predict storage time); and (4) description of a complex temporal process (for deciding on harvesting grapes).