Study on Site-Specific Earthquake Ground Motion Evaluation Models Utilizing Machine Learning Considering Epicentral Directions and Response Duration Times

Abstract

Aiming to acquire new knowledge by seismic motion evaluation from a new viewpoint, it was tried to create a site-specific earthquake ground motion evaluation model by machine learning using the earthquake ground motion records obtained in the past as learning data. The epicentral direction and the response duration time of the earthquake ground motion, which were not handled by the conventional attenuation relations formula, were also examined. As a whole, the observed values are well evaluated and modeled, within double to half of the observed values. The degree of influence of the epicentral direction on the response duration time is large, and in many cases, it is almost equal to or greater than each parameter of the conventional ground motion prediction formula.