Recurrent Convolutional Neural Networks help to predict location of Earthquakes
Abstract
We examine the applicability of modern neural network architectures to the midterm prediction of earthquakes. Our data-based classification model aims to predict if an earthquake with the magnitude above a threshold takes place at a given area of size 10 × 10 kilometers in 10-60 days from a given moment. Our deep neural network model has a recurrent part (LSTM) that accounts for time dependencies between earthquakes and a convolutional part that accounts for spatial dependencies. Obtained results show that neural networks-based models beat baseline feature-based models that also account for spatio-temporal dependencies between different earthquakes. For historical data on Japan earthquakes our model predicts occurrence of an earthquake in 10 to 60 days from a given moment with magnitude Mc > 5 with quality metrics ROC AUC 0.975 and PR AUC 0.0890, making 1.18 · 103 correct predictions, while missing 2.09 · 103 earthquakes and making 192 · 103 false alarms. The baseline approach has similar ROC AUC 0.992, number of correct predictions 1.19 · 103, and missing 2.07 · 103 earthquakes, but significantly worse PR AUC 0.00911, and number of false alarms 1004 · 103.