In　order　to　explore　the　mechanism　of　influencing　factors　of　the　electric　bike　(e-bike)speeding　behavior,the　global　positioning　system(GPS)trajectory　data　of　shared　e-bikes　is　used　to　realize　identification　and　risk　classification　of　speeding　behavior.　Considering　characteristics　such　as　land　use,roads,and　traffic　status,a　model　that　identifies　speeding　risk　on　road　segments　for　shared　e-bikes　is　created　based　on　machine　learning　algorithms.　Then,a　partial　dependency　plot　is　employed　to　analyze　the　influence　of　each　influencing　factor　on　speeding　risk　on　road　segments.　The　results　show　that　the　CatBoost　is　better　for　speeding　risk　identification　on　road　segments　than　the　random　forest　model.　As　land　use　density　and　curb　parking　density　decrease　and　bus　line　density,road　level,sidewalk　width,and　non-motorized　lane　width　increase,the　speeding　risk　on　road　segments　for　shared　e-bikes　increases.　In　addition,one-way　roads,non-physically　separated　sidewalks,　non-physically　separated　non-motorized　lanes,and　non-peak　hours　are　positively　associated　with　speeding　risk　on　road　segments.　This　study　provides　a　novel　method　for　identifying　and　analyzing　risky　e-bike　behavior　and　technical　support　for　non-motorized　traffic　safety　management.　©　2024　Southeast　University.　All　rights　reserved.