A chemometric study on the stream sediments of Meshkinshahr, NW Iran, using supervised and unsupervised classification methods

Selected multivariate data analysis methods have been used to study chemical features (some minor elements) of surface stream sediments collected from Meshkinshahr, NW Iran, for a geochemical investigation. A total of 630 samples were collected and levels of 22 elements (Au, Hg, Cr, Cu, Mn, Ni, Sr, Zn, Ba, Be, Ti, Ag, As, B, Bi, Co, Mo, Pb, Sb, Se, Sn, and W) were determined in each sample using fire assay for Au and inductively coupled plasma optical emission spectrometry (ICP-OES) for the rest. The data were processed using data-handling tools. Principal component analysis (PCA) and hierarchical cluster analysis (HCA) as unsupervised ordination and clustering methods monitored differences in samples according to the chemical compositions. PCA and HCA were used to express the relationships of the minor element distributions to identify the possible sources of these elements in that area. The results of PCA and HCA classified the samples into three main groups. Partial least squares discriminant analysis (PLS-DA) and support vector machine discriminant analysis (SVM-DA) were implemented as supervised classification models. PLS-DA and SVM-DA were applied to detect the relationship between the chemical contents and to discriminate different samples. Respectively, 545 and 85 samples were selected as calibration and prediction sets for classification methods. The prediction abilities of PLS-DA and SVM-DA models were 91.8 % and 79.0 %, respectively.