Methods of group classification based on the theory of multisets in the problem of localizing zones with different levels of seismic activity during mining.

The work is dedicated to assessing the applicability of supervised group classification methods developed on the basis of multiset theory for solving the problem of identifying zones with different degrees of seismic activity (using the example of one of the sections of the highly stressed rock massif of the Kukisvumchorr apatite-nepheline deposit). The initial objects for classification procedures are spatial cells into which the fieldis divided. Each spatial cell is described by a certain set of factors that, according to experts, have an impact on the occurrence of seismic events in a given cell. An original representation of spatial cells (their groups) as a set of multisets is proposed. Studies have been carried out aimed at identifying the influence of various options for presenting the initial data on the result of classification procedures. Representation of objects described by quantitative and / or qualitative features and existing in several versions (copies) in the form of multisets makes it possible not to transform qualitative features into numerical ones when performing clustering procedures and use methods of group classification of objects. Generalized decision rules of group classification for assigning objects (spatial cells) to four classes of seismic hazard are obtained. In contrast to the currently widely used technologies based on the neural network approach, in this work, the training result is not a “black box” in the form of a trained neural network, but a set of rules that can be easily interpreted, which increases the confidence of end users in decision-making procedures.