Earthquake prediction based on Bi-LSTM+CRF model and Spatio-Temporal Data

The destructiveness of earthquakes is self-evident. An earthquake not only brought irreversible disasters to the local people, but also brought great economic losses to the country. Therefore, earthquake prediction is a very necessary task. This study uses multiple Bi-LSTM models to predict different information of an earthquake from multiple dimensions: including time information, latitude and longitude information, magnitude information, etc., and finally integrated them together, so as to avoid displaying only one dimension of earthquake information. It can show the earthquake situation more comprehensively. For every two adjacent earthquake information, we use the method of taking the difference to train and predict the model, so that the model can extract the law between the data more correctly.