Traffic　safety　is　a　main　challenge　for　the　sustainable　development　of　urban　transportation.　Reasonable　coupling　classification　of　safety　and　driving　behavior　is　conducive　to　the　improvement　of　drivers＇　self-management　and　urban　traffic　problems　solving.　The　objective　of　this　paper　is　to　establish　a　method　based　on　driving　behavior　to　judge　the　degree　of　driving　safety,　which　is　used　to　mine　the　relationship　between　different　driving　behavior　and　driving　safety.　All　the　data　was　gathered　from　the　Beijing　taxi　drivers＇　driving　behavior　database　acquired　by　OBD　equipment.　Drivers　are　classified　into　three　categories:　safety,　general　safety,　and　danger　by　cluster　analysis.　In　this　paper,　the　graph　theory　is　adopted　for　data　coding　and　the　graph　of　driving　behavior　characteristics　based　on　micro　driving　behavior　indicators　is　established.　It　is　of　great　significance　to　provide　a　reference　and　basis　for　optimizing　drivers＇　driving　behavior　on　the　security　level.　©　2020　ASCE.