Machine learning for earthquake prediction

Abstract

Earthquake prediction for the region of Sendai, Japan was carried out in this study by using 7 seismic features as inputs to an artificial neural network. The seismic indicators are selected based on well-known seismological and geophysical facts and are able to represent the seismic state of a specific region. Prediction results were evaluated using 5 measures: Sn, Sp, P1, P0 and Accuracy. Overall, the prediction accuracy is not satisfactory, suggesting the use of a deeper learning neural network.