Pattern recognition components analysis

Nowadays, generating huge amounts of data is relatively simple. That means Data Reduction and Interpretation using multivariate statistical tools (chemometrics), such as pattern recognition, factor analysis, and principal components analysis, can be critically important to extracting useful information from the data. These subjects have been introduced in Chapters 5 and 6. 

J.M. Deane, Data reduction using principal components analysis. In Multivariate Pattern Recognition in Chemometrics, R. Brereton (Ed.). Chapter 5, Elsevier, Amsterdam, 1992, pp. 125-165. 

While principal components models are used mostly in an unsupervised or exploratory mode, models based on canonical variates are often applied in a supervisory way for the prediction of biological activities from chemical, physicochemical or other biological parameters. In this section we discuss briefly the methods of linear discriminant analysis (LDA) and canonical correlation analysis (CCA). Although there has been an early awareness of these methods in QSAR [7,50], they have not been widely accepted. More recently they have been superseded by the successful introduction of partial least squares analysis (PLS) in QSAR. Nevertheless, the early pattern recognition techniques have prepared the minds for the introduction of modem chemometric approaches. 

Two examples of unsupervised classical pattern recognition methods are hierarchical cluster analysis (HCA) and principal components analysis (PCA). Unsupervised methods attempt to discover natural clusters within data sets. Both HCA and PCA cluster data. 

Because protein ROA spectra contain bands characteristic of loops and turns in addition to bands characteristic of secondary structure, they should provide information on the overall three-dimensional solution structure. We are developing a pattern recognition program, based on principal component analysis (PCA), to identify protein folds from ROA spectral band patterns (Blanch etal., 2002b). The method is similar to one developed for the determination of the structure of proteins from VCD (Pancoska etal., 1991) and UVCD (Venyaminov and Yang, 1996) spectra, but is expected to provide enhanced discrimination between different structural types since protein ROA spectra contain many more structure-sensitive bands than do either VCD or UVCD. From the ROA spectral data, the PCA program calculates a set of subspectra that serve as basis functions, the algebraic combination of which with appropriate expansion coefficients can be used to reconstruct any member of the... 

Johnson, K.J., Synovec, R.E. (2002). Pattern recognition of jet fuels comprehensive GC x GC with ANOVA-based feature selection and principal component analysis Chemom. Intell. Lab. Syst. 60, 225-237. 

Keywords electronic nose principal component analysis pattern recognition chemical sensors sensor arrays olfaction system multivariate data analysis. 

The hypothesis of a normal distribution is a strong limitation that should be always kept in mind when PCA is used. In electronic nose experiments, samples are usually extracted from more than one class, and it is not always that the totality of measurements results in a normally distributed data set. Nonetheless, PCA is frequently used to analyze electronic nose data. Due to the high correlation normally shown by electronic nose sensors, PCA allows a visual display of electronic nose data in either 2D or 3D plots. Higher order methods were proposed and studied to solve pattern recognition problems in other application fields. It is worth mentioning here the Independent Component Analysis (ICA) that has been applied successfully in image and sound analysis problems [18]. Recently ICA was also applied to process electronic nose data results as a powerful pre-processor of data [19]. 

Software Availability. The SIMCA software is available in two forms, both developed by Wold (25) 1) an interactive, Fortran version which runs on Control Data Corporation (CDC) machines. The second set of programs are an interactive microcomputer version, SIMCA-3B, are available from Principal Data Components, 2505 Shepard Blvd., Columbia, MO 65201. The SIMCA-3B pattern recognition programs includes the CPLS-2 program used for PLS analysis and are available for CP/M (Digital Research, Pacific Grove, CA) and MS-DOS (Microsoft Corporation, Bellueve, WA) for 8088 or 80 86 based microcomputers. 

The SIMCA approach can be applied in all of the four levels of pattern recognition. We focus on its use to describe complex mixtures graphically, and on its utility in quality control. This approach was selected for the tasks of developing a quality control program and evaluating similarities in samples of various types. Principal components analysis has proven to be well suited for evaluating data from capillary gas chromatographic (GC) analyses (6-8). 

These applications demonstrate that pattern recognition techniques based on principal components may be effectively used to character zate complex environmental residues. In comparisons of PCBs in bird eggs collected from different regions, we demonstrated through the use of SIHCA that the profiles in samples from a relatively clean area differed in concentration and composition from profiles in samples from a more highly contaminated region. Quality control can be evaluated by the proximity of replicate analysis of samples in principal components plots. 

---