In engineering applications, attenuation relationship is one of the key elements in a seismic hazard analysis. However, it is difficult to find the optimal solution of coefficients in attenuation model using traditional mathematical methods because of its nonlinearity. Besides, it is not reasonable to use unweighted regression analysis in which each recording carried an equal weight due to nonuniform distribution of data with respect to distance. In this study, least squares method (LSM) and genetic algorithm (GA) are employed as nonlinear regression methods to find the optimal solution of parameters in attenuation model to compare the robustness and predicted accuracy of the two methods. Different weights (equal and unequal weights) of each recording were used to compare the adaptability of the weights for practical application. Considering the hazard and energy radiation of an earthquake, the unequal weight of each recording are defined as function of distance. Moreover, the regression analysis of horizontal peak ground acceleration (PGA) attenuation model in southwest Taiwan was studied.
