Estimation of the Peak Ground Acceleration using support vector machine and neural radius-based function network models

Prediction of the ground strong motion parameters is one way to evaluate the various earthquakes and to determine the amount of risk in each area which plays an important role in the evaluation of earthquake effects on the engineering projects design. In this study, the support vector machine (SVM) and neural radius-based function (RBF) network models as new artificial intelligence techniques were used to estimate the peak ground acceleration (PGA). For this purpose, the seismic parameters such as the magnitude, epicentral distance, focal depth, earthquake intensity were applied as input parameters of proposed models. Evaluation of obtained results for the estimation of PGA using the SVM and RBF models with empirical attenuation relationships and regression methods indicated that the presented SVM and RBF models can establish an appropriated relationship between the observed and calculated PGA values. Also, proposed models have more accuracy than classical approaches. The determination coefficient is 0.996 and 0.997 for SVM and RBF models, respectively where as the determination coefficient is 0.790 and 0.153 for linear regression and nonlinear regression, respectively.