Morgan P. Moschetti Eric M. Thompson Kyle Withers 2020 <div class="article-section__content en main"><p>We use a machine learning approach to build a ground motion model (GMM) from a synthetic database of ground motions extracted from the Southern California CyberShake study. An artificial neural network is used to find the optimal weights that best fit the target data (without overfitting), with input parameters chosen to match that of state-of-the-art GMMs. We validate our synthetic-based GMM with empirically based GMMs derived from the globally based Next Generation Attenuation West2 data set, finding near-zero median residuals and similar amplitude and trends (with period) of total variability. Additionally, we find that the artificial neural network GMM has similar bias and variability to empirical GMMs from records of the recent<span>&nbsp;</span><img class="section_image" src="https://agupubs.onlinelibrary.wiley.com/cms/asset/3b5d00e5-0a30-4f89-b352-4e759f0e46f2/grl60306-math-0001.png" alt="urn:x-wiley:grl:media:grl60306:grl60306-math-0001" data-mce-src="https://agupubs.onlinelibrary.wiley.com/cms/asset/3b5d00e5-0a30-4f89-b352-4e759f0e46f2/grl60306-math-0001.png"><span>&nbsp;</span>Ridgecrest event, which neither GMM has included in its formulation. As simulations continue to better model broadband ground motions, machine learning provides a way to utilize the vast amount of synthetically generated data and guide future parameterization of GMMs.</p></div> application/pdf 10.1029/2019GL086690 en American Geophysical Union A machine learning approach to developing ground motion models from simulated ground motions article