The main objectives of this article are to develop a technique to find source models that allow one to replicate observed strong ground motion records and to extrapolate strong ground motion synthesis to locations where strong motion was not recorded. A technique including the well known empirical Green’s function (EGF) approach along with a genetic algorithm is used, which allows the optimization of differences between the synthesized and observed ground shakings. The technique used is performed by comparing the elastic response spectra of observed seismograms at two stations with those of simulated data using the EGF method incorporating recorded aftershocks taken at each station. Moreover, a genetic algorithm approach is used to reduce differences between the simulated and recorded data in the form of elastic response spectra by changing the input parameters in the admissible ranges. To validate the proposed approach the three components of strong motion recorded at other stations were synthesized incorporating the input parameters obtained at previous stations. A comparatively good match of the simulated and recorded response spectra confirms the ability of the proposed technique to generate synthetic seismograms with suitable elastic response spectra.