Methodological and Technical Solutions for the Implementation of Clustering Algorithms in the GeoBazaDannych System

Abstract

In this paper, from the perspective of creating and maintaining geological or geoecological models, methodological and technical issues, ways of developing the system GeoBazaDannych (GBD), expanding its functionality are considered. The new functionality provided by the inclusion of executable data mining modules of the Wolfram Mathematica computer algebra system into the GBD is noted. In particular, it illustrates tools for preparing benchmark geodata sets for validating, testing, and evaluating related neural network models. Examples on representative data sets illustrate options for choosing the best clustering algorithms. Examples of representative datasets illustrate the options for choosing the best clustering algorithms. A series of calculations illustrate the effects of choosing the number of clusters, the clustering method, and metrics. Separately, options are considered when clustering takes into account not only the coordinates of observation points, but also the values in them.