Nowadays,　drug　abuse　is　a　universal　phenomenon.　It　can　bring　a　huge　damage　to　the　human　body　and　cause　an　irreversible　result.　It　is　important　to　know　what　can　lead　to　the　abusing　so　that　it　can　be　prevented.　In　order　to　prevent　the　abusing　of　those　drugs,　it　is　necessary　to　figure　out　what　elements　make　people　abuse　the　drug　and　how　they　relative　to　the　abusing.　According　to　the　drug　consumption　data　from　the　UCI　database,　Big　Five　personality　traits　(NEO-FFI-R),　sensation　seeking,　impulsivity,　and　demographic　information　are　considered　to　be　the　relative　elements　of　the　abusing.　However,　how　they　affect　on　the　abusing　of　drugs　is　not　clear　so　they　cannot　predict　the　probability　of　a　person　whether　he　is　going　to　abuse　a　drug.　There　are　many　traditional　ways　to　analysis　the　data　based　on　scoring,　such　as　give　every　element　a　score　and　but　they　can　only　tell　an　inaccurate　predictive　value.　Machine　learning　is　very　hot　nowadays　because　of　its　strong　learning　ability,　high　efficiency,　and　high　accuracy.　In　this　paper,　we　build　models　for　accurate　prediction　of　drug-abusing　with　the　personality　traits　and　some　other　information,　based　on　logistic　regression,　decision　tree,　and　random　forest　separately.　We　find　out　that　the　sensation　of　respondents　and　the　country　which　they　are　from　is　the　most　important　factor　for　drug　abuse.　And　we　can　get　a　conclusion　that　drug　abuse　is　not　only　depending　on　a　person’s　inner　being,　but　also　affected　by　the　environment　they　lived　in.　©　2021,　Springer　Nature　Singapore　Pte　Ltd.