Unsupervised Classification Analysis

Unsupervised classification - Cluster analysis. Unsupervised classification or cluster analysis is a way to find natural patterns in a data set. There are no independent true answers that can guide the classification and we are therefore restricted to construct a set of criteria or general rules that can highlight the interesting patterns in a data set. A data set consists usually of a set of objects that each are characterised by a feature vector x. To find patterns it is important to establish to what degree vectors are similar to each other. Similarity is often defined using a distance metric between object i and j... 

Finally, one class of unsupervised methods is represented by self-organising maps (SOM), or Kohonen maps, named after the Finnish professor Teuvo Kohonen. A SOM is a type of artificial neural network that needs to be trained but does not require labelling of the input vectors. Examples of classification analysis by SOMs in biomedical IR and Raman spectroscopy are given in references. ... 

Another type of ANNs widely employed is represented by the Kohonen self organizing maps (SOMs), used for unsupervised exploratory analysis, and by the counterpropagation (CP) neural networks, used for nonlinear regression and classification (Marini, 2009). Also, these tools require a considerable number of objects to build reliable models and a severe validation. 

Covell DG, Wallqvist A, Rabow A et al. Molecular classification of cancer unsupervised selforganizing map analysis of gene expression microarray data. Mol Cancer Ther 2003 2 317-332. 

Pattern recognition can be classified according to several parameters. Below we discuss only the supervised/unsupervised dichotomy because it represents two different ways of analyzing hyperspectral data cubes. Unsupervised methods (cluster analysis) classify image pixels without calibration and with spectra only, in contrast to supervised classifications. Feature extraction methods [21] such as PCA or wavelet compression are often applied before cluster analysis. 

Classification, or the division of data into groups, methods can be broadly of two types supervised and unsupervised. The primary difference is that prior information about classes into which the data fall is known and representative samples from these classes are available for supervised methods. The supervised and unsupervised approaches loosely lend themselves into problems that have prior hypotheses and those in which discovery of the classes of data may be needed, respectively. The division is purely for organization purposes in many applications, a combination of both methods can be very powerful. In general, biomedical data analysis will require multiple spectral features and will have stochastic variations. Hence, the field of statistical pattern recognition [88] is of primary importance and we use the term recognition with our learning and classification method descriptions below. 

The answers to these questions will usually be given by so-called unsupervised learning or unsupervised pattern recognition methods. These methods may also be called grouping methods or automatic classification methods because they search for classes of similar objects (see cluster analysis) or classes of similar features (see correlation analysis, principal components analysis, factor analysis). 

In complex systems where the number of groups to be separated during classification becomes larger, the performance of simple unsupervised methods (Section 3) degrades, requiring the use of more sophisticated supervised chemometric techniques. Additionally, in fields such a process NMR where there is a need for quantifying a component, the use of supervised methods becomes necessary. The different supervised methods described in the sections below have all been utilized in the chemometric analysis of NMR data for classification and/or quantitation. Examples utilizing these different techniques are discussed in Section 5. 

In unsupervised classification, the grouping of the objects is not known beforehand and finding such groupings is the purpose of the analysis. Classical methods to perform unsupervised classification include variants of cluster analysis [Mardia et al. 1979],... 

In chemometrics we are very often dealing not with individual signals, but with sets of signals. Sets of signals are used for calibration purposes, to solve supervised and unsupervised classification problems, in mixture analysis etc. All these chemometrical techniques require uniform data presentation. The data must be organized in a matrix, i.e. for the different objects the same variables have to be considered or, in other words, each object is characterized by a signal (e.g. a spectrum). Only if all the objects are represented in the same parameter space, it is possible to apply chemometrics techniques to compare them. Consider for instance two samples of mineral water. If for one of them, the calcium and sulphate concentrations are measured, but for the second one, the pH values and the PAH s concentrations are available, there is no way of comparing these two samples. This comparison can only be done in the case, when for both samples the same sets of measurements are performed, e.g. for both samples, the pH values, and the calcium and sulphate concentrations are determined. Only in that case, each sample can be represented as a point in the three-dimensional parameter space and their mutual distances can be considered measures of similarity. 

---