Inverse post-hoc problem of the cluster analysis with application to data discrimination

About the authors:

Sergey V. Dronov, Cand. Sci. (Phys.-Math.), Associate Professor, Department of Mathematics and Information Technologies, Altai State University (Barnaul)
Antonina S. Sazonova, Post-graduate student, Department of Mathematics and Information Technologies, Altai State University (Barnaul)

Abstract:

We consider the problem of informational importance of characteristics determination for a set of clusterized objects. These characteristics can be either numerical ones or non-numerical with categories. Using priori information about the natural order of clusters, we propose the way to range characteristics with respect to the degree of their importance. Assuming the discrimination function separating correctly the objects into the available clusters, we developed a new algorithm. The algorithm defines a type of the proper f_X transformation for each X characteristic. In this case, if we replace the discriminatory function X with f_X, then a new discrimination function will show the influence of the characteristic on cluster structure of the set of objects in the optimal way.

References:

1. Dronov, S.V., Gerasimova, A.S. On the problem of digitization of a cluster variable. Tr. vseross. molodezhnoi shkoly-seminara «Analiz, geometriia i topologiia» [Proceedings of Analysis, Geometry and Topology National Youth Workshop]. Barnaul, 2013. Pp. 54-58.

(in Russian).

2. Gerasimova, A.S. Clusterization of the objects with non-numerical features. Sovremennaia shkola Rossii. Voprosy modernizatsii: M-ly III Mezhdunar. nauch.-praktich. konf. [Modern school of Russia. Modernization issues]. Moscow, 2013. Pp. 6–9. (in Russian).

3. Gerasimova, A.S. Clusterization of the objects with non-numerical features and its use for the estimation of their connection strength. Izvestiia Altaiskogo gosudarstvennogo universiteta — Proceedings of Altai State University. 2013. Issue 1/2 (77). Pp. 66–69. (in Russian).

4. Aivazian, S.A., Bukhshtaber, V.M., Eniukov, I.S., Meshalkin, L.D. Prikladnaia statistika: Klassifikatsiia i snizhenie razmernosti [Applied statistics: classification and reduction of dimension]. Moscow, 1989. 607 p. (in Russian).

5. Dronov, S.V. Mnogomernyi statisticheskii analiz: Uchebnoe posobie [Multidimensional statistical analysis: textbook]. Barnaul, 2006. 221 p. (in Russian).