DD-SIMCA for 3-way data

Data Driven SIMCA can also be used for developing one-class classification models based on 3-way data, e.g. excitation-emission profiles, GC-MS, and similar. In this case, a 3-way decomposition (PARAFAC or Tucker) is used to fit the training set data and compute distances, parameters, etc.

This section describes how to do this using a simple simulated dataset. It assumes that you already know how conventional 2-way implementation of DD-SIMCA (ddsimca) works (as many plots and methods are similar) and understand the basics of PARAFAC and Tucker decompositions. We will describe the PARAFAC version (ddsimca.parafac()) in detail, and then show what is different in the Tucker version (ddsimca.tucker()).