Chemometrics feature selection

The success of pattern recognition techniques can frequently be enhanced or simplified by suitable prior treatment of the analytical data, and feature selection and feature extraction are important stages in chemometrics. Feature selection refers to identifying and selecting those features present in the analytical data which are believed to be important to the success of calibration or pattern recognition. Techniques commonly used include differentiation, integration, and peak identification. Feature extraction, on the other hand, changes... 

Feature selection, the choice of variables or features (combinations and transformations of original variables) relevant to the problem, is one of the most important aims of chemometrics. Reasons for feature selection are ... 

In this chapter we have been able to discuss only some of the more common and basic methods of feature selection and extraction. This area is a major subject of active research in chemometrics. The effectiveness of subsequent data processing and interpretation is largely governed by how well our analytical data have been summarized by these methods. The interested reader is encouraged to study the many specialist texts and journals available to appreciate the wide breadth of study associated with this subject. 

R. Leardi, R. Boggia, M. Terrile, Genetic Algorithms as a Strategy for Feature Selection, Journal of Chemometrics, 6 (1992), 267-281. 

R. Leardi, Application of a Genetic Algorithm to Feature Selection Under Full Validation Conditions and to Outlier Detection, Journal of Chemometrics, 8 (1994). 65-79. 

Johnson, K.J. Synovec, R.E. (2002). Pattern recognition of jet fuels comprehensive GCxGC with ANOVA-based feature selection and principal component analysis. Chemometrics and Intelligent Laboratory Systems, Vol.60, No.1-2, (January 2002), pp. 225-237, ISSN 0169-7439... 

Sinkov, N.A. Harynuk, J.J. (2011a). Cluster resolution A metric for automated, objective and optimized feature selection in chemometric modeling. Talanta, Vol.83, No.4, (January 2011), pp. 1079-1087, ISSN 0039-9140... 

Elliott GN, et al. Soil differentiation using flngerprint Fourier transform infrared spectroscopy, chemometrics and genetic algorithm-based feature selection. Soil Biol and Biochem 2007 39 2888-2896. 

R. Leardi. 2000. Application of genetic algorithm-PLS for feature selection in spectral data sets. Journal of Chemometrics 14 643-655. 

Decompositions of three-way arrays into these two different models require different data analysis methods therefore, finding out if the internal structure of a three-way data set is trilinear or nontrilinear is essential to ensure the selection of a suitable chemometric method. In the previous paragraphs, the concept of trilinearity was tackled as an exclusively mathematical problem. However, the chemical information is often enough to determine whether a three-way data set presents this feature. How to link chemical knowledge with the mathematical structure of a three-way data set can be easily illustrated with a real example. 

To access this information, go to www.spectroscopynow.com and select the Cherno-metrics channel. A website for the book is available - you should be able to access diis eidrer via the Features on the opening page or the left-hand side Education menu. If in doubt, use the search facility to find the book, or send an e-mail to chemometrics wiley.co.uk. 

In the second case, we took into account over 400 quinolones reported in the literature. Again, a chemometric approach based on multivariate characterization and design in the resulting latent variables permitted us to select a set of 32 molecules with a well-balanced structural variation on which to derive the QSAR models. Linear PLS modeling allowed ranking of the relative importance of individual structural features, and, by CARSO analysis, a new class of compounds was predicted, the lead of which was tested and shown to be as active as expected. This preliminary lead, after a proper modification, is presently being tested for further development. 

In addition to the conventional spectral analysis methods and chemometrics, two-dimensional (2D) correlation spectroscopy has recently been introduced to NIR spectroscopy (4,12-16). In this method spectral peaks are spread over a second dimension to simplify the visualization of complex spectra consisting of many overlapped bands and to explore correlation between the bands. There are two kinds of 2D correlation spectroscopy used in NIR spectroscopy. One is statistical 2D correlation proposed originally by Barton et al (16). This method employs cross-correlation based on the least-squares linear regression analysis to assess spectral changes in two regions, such as the NIR and mid-IR regions, that arise from variations in sample composition (16). In another 2D correlation spectroscopy proposed by Noda (12, 13), 2D spectra are constructed from a set of spectral data collected from a system under an external physical perturbation, which induces selective alterations in spectral features. 

---