Attention-Based Bi-Directional Long-Short Term Memory Network for Earthquake Prediction

An earthquake is a tremor felt on the surface of earth created by movement major pieces its outer shell. Till now, many attempts have been made to forecast earthquakes, which saw some success, but these attempted models are specific region. In this paper, an occurrence and location prediction model proposed. After reviewing literature, long short-term memory (LSTM) found be good option for building because memory-keeping ability. Using Keras tuner, best was selected from candidate models, composed combinations various LSTM architectures dense layers. This used seismic indicators catalog Bangladesh as features predict earthquakes following month. Attention mechanism added architecture improve model's accuracy, 74.67%. Additionally, regression built using layers epicenter distance predefined location, provided root mean square error 1.25.