Time series prediction of earthquake waves using RNN

Interests in earthquakes in Republic of Korea has been increased with recent large earthquakes. We used recursive neural networks to predict earthquake waves from early arrivals. Based on earthquake data generated in Korea, we trained multilayer RNN models using GRU, LSTM, and Bidirectional RNN. We adjusted hyperparameters through feature engineering. Then we compared performance of multilayer RNN models and compared the predicted time series with the recorded time series.