Machine Learning-Based Ground Peak Acceleration Attenuation Prediction Model.

For earthquake mitigation and disaster management, it is important to rapidly predict modified Mercalli intensity (MMI) attenuation patterns across the entire earthquake-affected area. The quick prediction of hazardous areas after a destructive earthquake can effectively reduce the number of casualties and extent of property damage. Peak ground acceleration (PGA) is closely related to earthquake MMI. Traditional PGA predictions are constructed based on empirical formulas, which struggle to explain the nonlinear relation between PGA and epicentral distance. To address this issue, we propose a machine learning-based PGA attenuation model. We utilized seismic data from the Japan K-NET and KiK-net networks collected during 2000–2023 to construct a dataset and applied three different models to fit the data. The best-fit model was selected based on the performance results of these models. Using this model, three earthquake events with MJMA >6 in 2023 were predicted to evaluate potentially hazardous areas resulting from earthquakes. The results indicated that the model could effectively predict potentially hazardous areas and rapidly generate seismic MMI maps after an earthquake. Additionally, the machine learning model successfully addressed the nonlinear relation between PGA and epicentral distance. [ABSTRACT FROM AUTHOR]