Exploring the unsupervised classification of seismic events of Cotopaxi volcano

This paper explores the use of six different clustering-based methods to classify long-period and volcano-tectonic seismic events and to find possible overlapping signals of non-volcanic origin that could occur at the same time or immediately after the occurrence of volcano-seismic events. According to the explored classifiers space, the BIRCH method with k = 2 was chosen as the best model in the classification of both pure seismic events, reaching a weighted balanced accuracy and accuracy scores of 0.81 and 0.88, respectively. The accuracy result represents a satisfactory and competitive classification performance when compared to the state of the art methods. Besides, the spectral-clustering method with k = 3 was able to classify seismic events with and without overlapped signals of non-volcanic origin, attaining a weighted balanced accuracy score of 0.51. This result was at least 0.18 units higher than the other classifiers. Additionally, the obtained true positive rates of 0.94 corroborated the excellent performance of this classifier to detect seismic events with overlapping. According to the obtained results, it is possible to state that the proposed clustering-based exploration was effective in providing competitive models for both the classification of uncontaminated seismic events as well as for the detection of seismic events with overlapped signals.