Generación de señales volcánicas artificiales de tipo LP (long-period) y VT (volcano-tectonic) a partir de una base de datos del volcán Cotopaxi usando la técnica de bootstrapping.

Abstract

The seismological study of a volcano is of vital importance since it provides ample information about its behavior and allows an exhaustive analysis of its internal volcanic activity. The permanent monitoring of a volcano allows estimating and predicting future events, and it is for this reason that an extraction of the most relevant characteristics of seismic events is necessary to activate or create prevention strategies in the face of catastrophes of this nature. Due to the little processed information that is available, it is proposed to use the bootstrapping technique to generate artificial events with similar characteristics and thus increase the amount of new information available. The information generated is very useful since it is generated based on the required research needs, for example, a type of important volcanic event is a VT (Volcano-Tectonic); however, very few such labeled events are available considering that machine learning algorithms typically require thousands of examples for the training process. To apply the bootstrapping technique, the original signals are previously prepared in the preprocessing stage. The bootstrap is an extremely powerful and widely applicable statistical tool that can be used to determine the uncertainty associated with a given statistical learning method. The similarity of the generated signals is evaluated by using the Frechet Distance (FD).