Early　identification　of　potential　disruptive　technologies　is　critical　to　corporate　R&D　investment　decisions　and　government　R&D　strategy　decisions.　However,　how　early　identifying　potential　disruptive　technologies　has　been　the　focus　of　academic　community.　Therefore,　in　this　paper,　we　propose　a　framework　for　early　identification　of　potential　disruptive　technologies　based　on　machine　learning　and　text　mining.　In　the　framework,　we　firstly　obtain　relationships　pattern　between　the　characteristics　of　highly　cited　papers　and　their　citation　trends　by　using　machine　learning　models.　Then,　we　use　the　relationships　pattern　to　identify　potential　highly　cited　papers,　solving　the　time　lag　problem　of　using　citation　analysis　to　identify　potential　highly　cited　papers.　Secondly,　we　construct　a　breakthrough　index　based　on　breakthrough　research　characteristics　to　identify　breakthrough　papers　from　potential　highly　cited　papers.　Finally,　we　use　text　mining　methods　to　obtain　breakthrough　research　topics　from　breakthrough　papers,　and　identify　potential　disruptive　technologies　by　analyzing　and　evaluating　breakthrough　research　topics.　An　empirical　study　was　conducted　in　the　field　of　chemistry　discipline　to　verify　the　framework＇s　feasibility　and　effectiveness.　This　paper　provides　a　new　perspective　for　the　early　identification　of　potential　disruptive　technologies　and　lessons　for　breakthrough　research　identification　and　evaluation.　　©　2023　PICMET　(Portland　International　Center　for　Management　of　Engineering　and　Technology(.