Concrete　Filled　Steel　Tube　(CFST)　is　favorable　to　utilize　in　the　construction　of　arch　bridges.　Since　CFST　are　mostly　intended　to　carry　compression　loads,　the　axial　compressive　capacity　is　of　primary　importance　in　the　of　CFST　columns.　Current　standards　limit　their　applicability　to　conventional　material　strength　and　geometric　dimensions　of　CFST,　sometimes　failing　to　meet　the　requirements　of　modern　arch　bridges.　This　study　has　developed　a　probabilistic　Machine　Learning　(ML)　model　based　on　the　NGBoost　algorithm.　The　results　demonstrate　that　the　NGBoost　model　with　a　LogNormal　distribution　provides　both　accurate　and　probabilistic　predictions,　surpassing　the　performance　of　the　XGBoost　model　and　the　NGBoost　model　with　a　Normal　distribution.　Furthermore,　we　have　employed　the　SHapley　Additive　exPlanations　(SHAP)　method　to　interpret　the　probabilistic　model.　It　was　revealed　that　the　column　dimensions　exert　the　most　significant　influence　on　the　axial　compressive　capacity　of　CFST.　©　The　Author(s),　under　exclusive　license　to　Springer　Nature　Switzerland　AG　2025.