Research On Earthquake Prediction Method Based On Deep Learning

Earthquake is a huge natural disaster,to the safety of human life and property has brought a great threat,to be able to predict the occurrence of earthquake in advance and do the corresponding measures is very necessary,so the earthquake prediction is a very important task.For a long time,experts of earthquake prediction has been exploring a more accurate way of earthquake prediction,they also produce a lot of schools,but these methods have the following defects: first of all,the seismic data acquisition is difficult,followed by the traditional method can only predict earthquakes a dimension of information,such as the magnitude or latitude and longitude,prediction information lack of completeness.In view of the above problems,this paper proposes a method that can predict the information of each dimension of earthquake,and uses the deep learning model to make the prediction.The main work of this paper is as follows:1.Firstly,this paper obtained seismic data from the USGS seismic data network to extract the required seismic data dimensions,including time,magnitude,latitude and longitude,and focal depth,etc.,and then stitched these feature information together and input it into multiple deep learning models: In BI-LSTM +CRF model,to predict the information of various dimensions of future earthquakes,different deep learning models are responsible for predicting the information of different dimensions of future earthquakes,and these information are finally integrated to obtain the information prediction of various dimensions of future earthquakes.2.According to the characteristics of seismic prediction task data,the BI-LSTM+CRF model was improved,and the MF-LSTM+CRF model was designed.The model improved the way of adding features,adding features to the LSTM nodes in the process of transverse information transmission through the gate mechanism.It not only reduces the influence of the added features on the model,but also makes the features integrate into the model more naturally,and greatly improves the prediction effect.In order to prove the effect of the model in this paper from the side,this paper chooses the previous research results,BP model,to carry out the comparison experiment.Since BP model is used to predict the magnitude of the earthquake,the experiment only compares the magnitude from one dimension.The results show that the prediction effect of the proposed model is more stable and better than that of the BP model in the range of each earthquake magnitude.