Early Estimation of Earthquake Magnitude Using Machine Learning

Abstract

Seismic parameters provide important information that describes the characteristics of an earthquake. The magnitude parameter is one of the essential seismic parameters in making the right decision regarding earthquake disaster mitigation. Determining the magnitude of an earthquake must be done early because this information represents the size of the earthquake and the potential damage it causes. If the determination of the earthquake’s magnitude is delayed, emergencies such as the evacuation of residents and post-disaster recovery may be disrupted. This study attempts to estimate the earthquake magnitude parameters based on Primary (P) wave signals using several machine learning algorithms for regression, such as Neural Network Regression (NNR), Random Forest Regression (RFR), and Support Vector Machine Regression (SVMR). The experimental results show that the RFR can produce the best estimation with an R-squared (R2) value of 0.946 and a root mean square error (RMSE) of 0.087.