Decision criteria for soft independent modelling of class analogy applied to near infrared data

SIMCA (soft independent modelling of class analogy) is a well known pattern recognition method which describes each class separately in a principal co mponents (PC) space. New objects are considered to belong to the class if t heir Euclidean distance towards the constructed PC space is not significant ly larger than the Euclidean distance of the class objects towards their PC space. The large number of wrongly rejected objects (alpha-error), which i s a known problem of SIMCA, was examined. Using scores, predicted by leave- one-out cross-validation, instead of the original scores, obtained after PC A on the class objects, to compute the distance towards the class model cle arly improves the decision criterion of the original SIMCA for a data set c onsisting of near infrared (NIR) spectra of tablets, The original SIMCA and modifications using different distance measures as defined by Hawkins (Mah alanobis distance) and Gnanadesikan were compared with respect to classific ation and their robustness towards the number of PCs selected to describe t he different classes. SIMCA modified with the Mahalanobis distance was foun d to be a good alternative of the original SIMCA which, for the presented N IR data set, seems to be more robust for finding outliers when the exact nu mber of PCs to build the model is not known. (C) 1999 Elsevier Science B.V. All rights reserved.