Support vector machines for classification of soils according to geographic origin based on their radionuclide content

Abstract

The paper introduces support vector machines (SVM), a recent method in statistical learning theory, used to recognize and classify soils according to their geographic origin. The classification was performed based on activities of seven radionuclides determined by gamma-ray spectrometry. The radionuclides of uranium and thorium series (226Ra, 232Th, 235U, 238U) and 40K were used to differentiate investigated areas based on geology, while cosmogenic beryllium ( 7 Be) and anthropogenic 137Cs were used to differentiate areas according to their susceptibility to fallout. The performances of the proposed method was compared to those of principal component analysis (PCA), linear discriminant analysis (LDA), k-nearest neighbours (kNN), soft independent modelling of class analogy (SIMCA) and artificial neural networks (ANN) applied to the same dataset.