H.A. Kravtsov

Èlektron. model. 2018, 38(6):15-24


The existing methods of classifier assessment use a set of classes which are comparable both by the probability of appearànce and by semantical interrelation that is they are semantically independent. The developed theory of calculus over classification permits solving the issue of classifier assessment for hierarchical classifications. This papper contains the example of calculation of the precision and completeness of classes of plane-level and multi-level classification with the same confusing matrix.


classification, classifier, semantic, precision, completeness, measure of difference.


1. Kravtsov, H.A. (2016), “Measure of difference between classifications”, Elektronnoe modelirovanie, Vol. 38, no. 4, pp. 81-97.
2. Kravtsov, H.A. (2016), “Model of computations on classifications”, Elektronnoe modelirovanie, Vol. 38, no. 1, pp. 73-87.
3. Flakh, P. (2015), Machinnoe obuchenie [Machine learning: The art and science of algorithms that make sense of data], Izd-vo “DMK Press”, Moscow, Russia.
4. Kokhonen, T. (2008), Samoorganizuyuschiesya karty [Self-organizing maps], Binom, Moscow, Russia.
5. Ushakov, D.N. (1935), Tolkovyi slovar russkogo yazyka [Russian definition dictionary], Sovetskaya entsiklopediya, Moscow, Russia.
6. Khaykin, S. (2006), Neyronnye seti. Polnyi kurs [Neural networks: A comprehensive foundation], Izd. “Vilyams”, Moscow, Russia.
7. Bazhenov, D. (2012), Otsenka klassifikatora. Tochnost, polnota, F- mera [Classification performance evaluation. Precision, completeness, F-measure], available at: (accessed 2016).
8. Kharman, H.H. (1972), Sovremennyi faktornyi analiz [Modern Factor Analysis], Statistika, Moscow, Russia.
9. Struhl, S. (2015), Practical text analytics: Interpreting text and unstructured data for business intelligence, 1st edition, Kogan Page, London, Philadelphia, New Delphi.

Full text: PDF (in Russian)