ISSN 2079-3537      

Scientific Visualization
Issue Year: 2017
Quarter: 4
Volume: 9
Number: 5
Pages: 86 - 104
Authors: A.E. Bondarev (Russian Federation), V.A. Galaktionov (Russian Federation), L.Z. Shapiro (Russian Federation)
The paper is recommended by program committee of International Conference «Visual Analytics»
Address: A.E. Bondarev
Keldysh Institute of Applied Mathematics RAS, Moscow, Russian Federation

V.A. Galaktionov
Keldysh Institute of Applied Mathematics RAS, Moscow, Russian Federation

L.Z. Shapiro
Keldysh Institute of Applied Mathematics RAS, Moscow, Russian Federation
Abstract: The paper considers the problems of visual analysis of multidimensional data sets. For visual analysis, the known approach of constructing elastic maps, described in [1-3], is used. Elastic maps are used as methods of mapping the initial data points of original data points mapping to enclosed manifolds having less dimensionality for the subsequent analysis of cluster structures in the original data volume. Diminishing the elasticity parameters, one can design map surface which approximates the multidimensional dataset in question much better. The points of dataset in question are projected onto the map. The extension of designed map to a flat plane with mapping into the space of the first three main components allows one to get an insight about the cluster structure of multidimensional dataset. The design of elastic cards does not require a priori information about the data and does not depend on the nature of the data, the origin of the data, etc., which is an important advantage of this method. The paper presents the results of applying elastic maps for the visual analysis of multidimensional data sets of various origins. In particular, the problem of the analysis of textual information represented in the form of a multidimensional array of frequencies of joint use of verbs and nouns is considered. Data processing techniques are described that allow improving the results obtained. For example, the application of the "quasi-Zoom" method makes it possible to significantly improve the results in the region of condensation of the points of the multidimensional space under study.
Language: Russian

Open Article   Download ZIP archive