Analysis of the Samples with an Unknown Matrix Using Data Mining Algorithms

Abstract

In determining a limited number of analytes in samples having a complex chemical composition with an unknown matrix, the combination of data mining algorithms (problems of clustering and regression) is proposed. This makes it possible to compensate for the influence of the components of the host medium on the intensity of the analytical line of an element being determined. The technology developed is tested in the X-ray fluorescence determination of S, Fe, Cu, Zn, and As in float concentrate samples during processing of polymetallic ores and V and Fe in synthetic film samples that are adequate in physicochemical properties to samples of welding fumes deposited on a filter. The error of the results of analysis has decreased by a factor 1.5–5 compared to the use of the Lucas-Tooth classical regression equation. The developed technology considerably increases the rapidity of analysis when it is used with X-ray spectrometers of consecutive action.