MATHEMATICAL MODELING

Analysis of data mining methods for classification of multidimensional medico-biological objects

G. M. Popova, V. N. Stepanov

V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow

Abstract: The paper presents the analysis and comparison of different methods for classification of multidimensional biomedical objects, applied in the field of "data mining": decision trees, support vector machines and hybrid neural networks. The effectiveness of these methods was checked by the example images of sputum cell structures on cytology preparations.

Keywords: Data mining, support vector machine, classification and regression trees, neuro-fuzzy networks, classification of multidimensional objects, cytological preparations.