Tetrastigma　hemsleyanum　Diels　et　Gilg　(T.　hemsleyanum),　also　known　as　Sanyeqing　in　Chinese,　is　a　rare　medicinal　herb.　The　quality　of　its　crude　drug　highly　depends　on　the　geographical　origin　of　the　plants.　The　traditional　supervised　discrimination　methods　based　on　near-infrared　spectroscopy　(NIRS)　construct　a　classifier　to　discriminate　samples　with　previously　trained　geographical　origins.　Those　methods　have　limited　power　since　a　lot　of　real　samples　are　unknown　and　untrained　by　the　classifier.　It　is　necessary　to　develop　a　discrimination　method　that　takes　untrained　geographical　origins　into　consideration.　In　this　study,　a　novel　identification　method　is　developed　using　an　improved　naive　Bayesian　classifier　combined　with　a　clustering　algorithm　by　fast　search　and　find　of　density　peaks　(INBC-CFSFDP).　In　detail,　the　INBC　is　proposed　to　distinguish　samples　with　new　geographical　origins　that　are　not　given　in　the　training　database　and　the　CFSFDP　is　used　to　further　identify　the　interior　categories　of　those　samples　with　new　geographical　origins.　The　experimental　results　show　that　regardless　of　whether　the　geographical　origins　are　indexed　in　the　training　database　or　not,　they　can　be　accurately　identified　using　an　INBC-CFSFDP,　and　they　demonstrate　that　the　proposed　method　is　easy,　effective,　fast,　and　feasible　to　discriminate　the　geographical　origin　of　T.　hemsleyanum.