Traffic　crashes　have　posed　a　serious　threat　to　global　health　and　have　had　a　negative　impact　on　social　and　economic　development.　In　2015,　1.25million　people　died　in　traffic　crashes　around　the　world,　over　90%　of　which　were　at　least　partially　caused　by　human　error.　This　research　intends　to　develop　a　new　driver　training　tool　which　can　help　reduce　human　error　by　correcting　drivers＇　risky　driving　behaviors.　On　the　basis　of　the　theory　of　planned　behavior　(TPB),　this　research　proposed　a　perception-norm-execution　(PNE)　driving　simulator-based　training　model.　To　evaluate　the　effectiveness　of　this　PNE　training　model,　five　risky　driving　behaviors　(i.e.　improper　lane　changes,　unsafe　overtaking,　red-light　running,　speeding,　and　distracted　driving)　were　selected　as　training　scenarios.　A　total　of　44　participants　were　recruited　to　participate　in　the　trainings.　Twenty-two　participants　received　the　PNE　model　(PM)　training,　and　the　remaining　participants　received　the　normal　(video-based)　model　(NM)　training.　This　research　used　the　empirical　data　collected　from　a　driving　simulator　to　evaluate　participants＇　driving　performance　before　and　after　they　received　the　training.　It　is　found　that　both　training　models　significantly　improved　driving　performance,　but　the　PM　training　was　demonstrated　to　be　more　effective　than　the　NM　training　in　terms　of　correcting　drivers＇　risky　driving　behaviors　such　as　improper　lane　changes,　unsafe　passing,　and　running　red　lights.　The　research　findings　can　be　used　to　change　policies　and　develop　more　effective　driver　education　and　training　programs.