Using a deterministic physics-based method, a simulated near-fault ground motion dataset by considering magnitude uncertainty, focal mechanism parameters, and focal depth was constructed, and the pulse-like ground motion and its parameters were obtained by using Baker's wavelet method. A pulse occurrence probability model with different components was established for various magnitudes and focal mechanisms, and it was further compared with Iervolino and Cornell's empirical prediction model. Meanwhile, a mathematical model was established for the spatial azimuth factor of pulse period, and the relationship between the spatial azimuth factor with the pulse-like station's location was discussed. The results show that the pulse period in the simulated near-fault ground motion dataset generally increases with the increase of magnitude, and the mean value of the pulse period under different magnitudes is close to the value predicted by the empirical model. The pulse area estimated by the pulse occurrence probability model in this paper also increases with the increase of magnitude. For strike-slip, the pulse area is mainly concentrated on both sides of the fault, while for dip-slip, the pulse area is a belt-like distribution with the direction parallel to the fault strike. The prediction results in this paper are generally consistent with Iervolino and Cornell's empirical prediction model. Especially for dip-slip, the predicted shape in this paper is more consistent with the distribution of the simulated PGV, while the pulse occurrence probability on both sides of the fault can be overestimated by the empirical model. Finally, both the magnitude and the pulse-like station's spatial location have a significant impact on the spatial azimuth factor. The established spatial azimuth factor of pulse period can be used to estimate the pulse period of horizontal components under different magnitudes.