There　are　many　factors　that　affect　the　compressive　strength　of　carbon　nanotubes/　cementious　composites.　However,　there　is　a　lack　of　comprehensive　research　on　the　effect　of　various　properties　of　CNTs　on　the　compressive　strength　of　cement-based　materials.　In　this　paper,　machine　learning　was　used　to　predict　the　compressive　strength　of　single　system　and　multi-system　of　cement-based　materials　mixed　with　CNTs,　and　the　optimal　parameters　of　CNTs　properties　were　also　determined.　Linear　Regression　(LR),　Support　Vector　Regression　(SVR),　Random　Forest　(RF)　and　Extreme　Gradient　Boosting　(XGB)　were　used　to　predict　the　compressive　strength　of　the　cement　paste,　mortar　and　concrete(single　system)　respectively.　The　results　showed　that　the　R2　of　prediction　of　RF　and　XGB　models　was　greater　than　0.9　for　all　three　systems,　and　the　accuracy　of　XGB　was　higher.　The　data　of　the　three　systems　were　combined　into　a　new　dataset(multi-system),　RF　and　XGB　were　used　for　its　prediction.　The　results　indicated　that　the　same　prediction　errors　were　achieved　by　RF　and　XGB,　and　the　R2　was　0.93.　SHAP　model　was　utilized　to　explain　the　prediction　results.　The　analysis　of　single　system　showed　that　the　content　and　diameter　of　CNTs　had　obvious　effects　for　all　the　three　systems,　and　the　influence　degree　of　CNTs　in　concrete　was　less　than　that　of　paste　and　mortar.　Some　optimal　parameters　of　CNTs　were　determined　by　the　multi-system　SHAP　results:　the　optimal　values　of　the　length　and　diameter　of　CNTs　were　20　mu　m　and　25　nm,　and　the　content　should　be　within　0.1　%.