Classification of Seismic Events Using Recurrent Neural Networks

Abstract

The main goal of this bachelor thesis is the study and implementation of recurrent neural networks for classifying types of seismic events. The thesis presents the basic theory of neural networks, the design of a custom three-layer recurrent neural network (LSTM type) and suitable preprocessing to accelerate the training of such a network. The achieved results are compared with the results achieved using the standard LSTM-FCN architecture on data from the OKC seismic station (Ostrava-Krásné Pole) from 2007–2022.