Machine Learning Implementation for Estimation of Earthquake Magnitude Using Strong-Motion Data

Determining earthquake parameters in a short time presents a trade-off between speed and accuracy. To enhance both the accuracy and speed of magnitude determination after the arrival of the P-wave, we introduce a machine learning-based magnitude prediction model using the Random Forest Regression method (RFmag) for earthquake magnitude estimation. In this study, strong-motion data obtained from the Agency for Meteorology Climatology and Geophysics (BMKG) were used to calculate the input parameters for the RFmag model. The model utilizes 12 significant parameters extracted from 30 seconds of strong-motion data recorded after the P-wave arrival as input features. The results showed that the RFmag model achieved an impressive accuracy rate of 94.49% with a standard deviation of 0.36, indicating its capability to estimate earthquake magnitude effectively. Therefore, the RFmag model demonstrates significant potential for application in real-time Tsunami Early Warning (TEW) operations, offering a more efficient and reliable method for rapid magnitude estimation in Indonesia's seismically active regions.