Premium
An evaluation of SIMCA. Part 1 — the reliability of the SIMCA pattern recognition method for a varying number of objects and features
Author(s) -
Dröge J. B. M.,
Van 'T Klooster H. A.
Publication year - 1987
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/cem.1180010404
Subject(s) - pattern recognition (psychology) , artificial intelligence , classifier (uml) , mathematics , reliability (semiconductor) , computer science , power (physics) , physics , quantum mechanics
The SIMCA pattern recognition method has been evaluated with pseudo random data sets. The number of objects varied from 5 to 50 and the number of features from 5 to 300. First, the determination of the significant number of PCs in the SIMCA models by the cross‐validation method was compared with the indicator function. The results showed that for the lower dimensions (≤ 15 objects or ≤ 15 features) the indicator function produces more reliable results. Second, the classification results with SIMCA were analysed for data sets with two equally sized classes and a varying number of objects and features, using the recall function as the evaluation criterion. The results showed that the SIMCA classifier produces reliable results at the first classification level, even for a low object/feature ratio (5/300). However, at the second level the classification performance of SIMCA decreases rapidly with an increasing number of features, even when the data set consists of two very well separated classes and little random error.