Application of constrained clustering methodsfor researching mining-induced seismicity

Abstract

The purpose of this study is to assess the applicability of constrained clustering methods for researching mining-induced seismicity. The paper proposes an original method of constrained clustering, which uses the hierarchical clustering scheme as a base (new clusters are created by combining smaller clusters). The distance between the centroids of the clusters is used as a meas-ure of the difference. The novelty of the method lies in the ability to take into account user con-straints on which objects must fall into one cluster, and which should not. In other words, within the framework of the proposed approach, when assigning objects to one or different clusters, not only the distances between the objects, but also the values of their attributes are analyzed. The proposed method allows stopping the clustering procedure in case of violation of user constraints. The studies are an initial attempt to assess the applicability of constrained clustering methods for researching mining-induced seismicity and to show a possibility in principle to assess the degree of mining and geological factors impact to seismicity. Various combinations of stationary and semi-stationary factors impacting to rock mass seismicity are investigated. The results of clustering can be used to pre-allocate areas with different seismic activity. Further research should be carried out taking into account not only the number of seismic events, but also their energy, the distance between them, etc. The use of constrained clustering methods allows us to speed up the computation process by reducing the number of alternatives at each clustering step. Guaranteed implementation of user restrictions within the framework of the proposed method increases the confidence of subject matter experts in clustering procedures. The obtained results make it possible to judge the prospects for the further development of limited clustering methods for the study of mining-induced seismicity.