Quantifying　the　near-fault　effect　and　establishing　a　reasonable　model　of　near-fault　pulse-like　ground　motions　are　particularly　important　for　seismic　design　of　structures　in　near-fault　regions.　Given　the　pronounced　randomness　associated　with　earthquakes,　this　study　first　proposes　a　novel　stochastic　model　of　near-fault　pulse-like　ground　motions　by　combining　the　improved　finite-fault　model　(IFFM)　and　the　multivariate　copula-based　velocity-pulse　model　(CVPM).　Further,　a　probability　density　evolution　method　(PDEM)　based　stochastic　simulation　method　is　developed,　by　which　the　model　parameters　can　be　determined　in　a　unified　probability　space　so　as　to　ensure　the　consistency　of　two　independent　models.　For　illustrative　purposes,　the　observed　records　collected　from　the　1999　Chi-Chi　earthquake　are　used　to　generate　new　stochastic　ground　motions　set.　Two　ground　motions　sets　based　on　classical　stochastic　simulation　methods　are　also　presented　for　comparison.　Numerical　results　show　that　the　proposed　method　for　stochastic　simulation　of　near-fault　pulse-like　ground　motions　is　reliable;　the　statistics　of　peak　ground　accelerations　and　spectral　characteristics　of　simulated　samples　are　consistent　with　station　records.　Besides,　the　proposed　method　accommodates　the　noteworthy　randomness　and　proportion　consistency　of　components　associated　with　near-fault　pulse-like　ground　motions,　making　it　suitable　for　the　stochastic　response　and　reliability　analysis　of　seismic　structures　in　near-fault　regions.　This　superiority　is　challenging　to　classical　stochastic　simulation　methods　that　lack　reasonable　consideration　of　randomness　and　correlation　associated　with　model　parameters.