ISSN:
0886-9383
Keywords:
Pattern recognition
;
SIMCA
;
PCA
;
Classification
;
Bacteria
;
Pyrolysis mass spectra
;
Recall
;
Cross-validation
;
Indicator function
;
Pseudo random data sets
;
Leave-x-out
;
Chemistry
;
Analytical Chemistry and Spectroscopy
Source:
Wiley InterScience Backfile Collection 1832-2000
Topics:
Chemistry and Pharmacology
Notes:
As part of a critical evaluation of the pattern recognition method SIMCA, three data sets containing pyrolysis mass spectra from bacteria were analysed using the SIMCA classifier. Each set consisted of two classes, Pseudomonas and Serratia bacteria, each class containing ten mass spectra and each mass spectrum having 285 spectral features.The results indicate that for these py-MS data sets, with low object/feature ratio, the SIMCA classifier produces satisfactory results at the first classification level. At the second level, however, the classification results are not reliable, even after deleting outliers. A comparison of the cross-validation method and Malinowski's indicator function for the determination of the number of significant principal components showed that the cross-validation method is less stable and therefore less reliable than the indicator function.
Additional Material:
11 Tab.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1002/cem.1180010405
Permalink