Earthquake Magnitude Estimation Based on Machine Learning: Application to Earthquake Early Warning System

Abstract Indonesia has high level of seismic activity, so determining magnitude an earthquake is important in the Earthquake Early Warning System. In System, parameter must be estimated earlier, that warnings can disseminated before S and surface waves arrive. previous studies machine learning technology used to recognized events extract hidden information with massive datasets. This study was a preliminary, proposed alternative methods calculate as fast possible, data 1s 3 seconds after P wave from 3-component single station raw seismogram historical developed classification deep neural network (DNN) model, classical random forest (RF) algorithm regression (DNN). Results statistical analysis show waveform modelled by models. Classification DNN Model we constructed reaches good pattern which final loss 0.63. If it benchmarked another model such Random (RF), better than RF determined RF. Our recommendation related estimate modelling are using larger dataset. our study, relatively small dataset, option. Another suggestion this work utilizing Regression DNN, resulting best estimation magnitude.