You can view the full text of this article for free using the link below.

Title: LSTM-based earthquake prediction: enhanced time feature and data representation

Authors: Asmae Berhich; Fatima-Zahra Belouadha; Mohammed Issam Kabbaj

Addresses: AMIPS Research Team, E3S Research Center, Ecole Mohammadia d'Ingénieurs, Mohammed V University in Rabat, 10090, Morocco ' AMIPS Research Team, E3S Research Center, Ecole Mohammadia d'Ingénieurs, Mohammed V University in Rabat, 10090, Morocco ' AMIPS Research Team, E3S Research Center, Ecole Mohammadia d'Ingénieurs, Mohammed V University in Rabat, 10090, Morocco

Abstract: In the last decades, many studies have emerged with different approaches in the field of earthquake prediction. Statistical and machine learning approaches have been used. However, these contributions remain immature. Some of them have not led to a successful prediction. Others have not been able to predict earthquakes so efficiently. Consequently, research into more relevant methods appropriates to this field will be important, as it would improve accuracy, performance, and dynamicity. This paper suggests applying the well-known deep learning algorithm long short-term memory to predict earthquakes in Moroccan regions. The features used in the prediction takes the most influencing and correlated datasets, it calculates an appropriate time feature that is simpler and more precise. The optimal hyperparameters values of our models are retrieved by the grid search technique. The performance of our model is compared with deep neural networks. The final results demonstrate that our model is more effective.

Keywords: earthquake prediction; deep learning; neural networks; LSTM; long short-term memory; seismic dataset; recurrent neural networks; magnitude prediction; hyperparameters optimisation; grid search.

DOI: 10.1504/IJHPSA.2021.115499

International Journal of High Performance Systems Architecture, 2021 Vol.10 No.1, pp.1 - 11

Received: 22 Aug 2019
Accepted: 19 Jun 2020

Published online: 07 Jun 2021 *

Full-text access for editors Full-text access for subscribers Free access Comment on this article