Based　on　three　years　of　maintenance　data　of　56　buses　with　three　types　of　fuel,　the　fault　situation　of　buses　has　been　analyzed.　By　using　the　regression　method　of　least　absolute　shrinkage　and　selection　operator　(LASSO),　the　main　influencing　factors　of　bus　fault　risk　were　detected　by　establishing　a　LASSO　model.　Five　machine　learning　methods　were　used　to　construct　the　bus　fault　risk　prediction　model,　and　the　prediction　accuracy　was　compared　and　analyzed.　According　to　the　prediction　model　results　and　corresponding　analysis,　the　bus　fault　risk　prediction　system　is　constructed.　Results　show　that　13　variables　with　the　most　significant　correlation　with　bus　fault　are　screened　out　by　LASSO　regression　model.　It　is　found　that　the　Bayesian　network　model　has　the　highest　prediction　accuracy　of　86%.　Furthermore,　through　this　model　analysis,　the　influences　of　various　influencing　factors　on　the　occurrence　of　bus　faults　can　be　obtained　more　intuitively.