A　passenger　evacuation　network　at　subway　stations　was　established,　the　degrees　of　network　nodes　were　analyzed　to　identify　key　nodes,　and　a　modified　walking　evacuation　network　was　proposed　after　node　failures.　A　time　prediction　model　for　passengers　traveling　through　the　gate　and　stair/escalator　nodes　was　designed　based　on　the　support　vector　regression　algorithm,　and　its　prediction　performance　was　analyzed.　A　quantitative　relationship　between　the　number　of　passengers　and　travel　time　was　obtained.　Three　objective　functions,　namely　the　total　evacuation　time　including　the　node　and　road　section　travel　time,　the　total　road　risk,　and　the　total　congestion　cost,　were　established.　A　multiobjective　route　optimization　model　for　passenger　evacuation　at　subway　stations　during　a　fire　outbreak　was　constructed,　and　a　solution　method　of　optimization　model　was　proposed　based　on　the　genetic　algorithm.　The　evacuation　movement　of　passengers　at　subway　stations　during　a　fire　outbreak　was　simulated,　and　the　evacuation　efficiencies　under　the　Pareto　solution　of　the　route　optimization　model　was　analyzed,　then　the　optimization　degree　of　the　route　optimization　strategy　was　evaluated.　A　WeChat　applet　for　passenger　dynamic　guidance　was　designed,　providing　a　possible　solution　for　the　timely　release　of　evacuation　route　recommendation　information.　Research　results　indicate　that　the　mean　absolute　error　of　the　passenger　travel　time　prediction　model　at　the　nodes　for　the　validation　set　can　be　as　low　as　0.000　375,　and　the　robust　indicator　value　is　up　to　0.999　334,　indicating　a　high　degree　of　consistency　between　predicted　data　and　real　data.　The　average　relative　error　between　the　field　collected　data　at　the　gate　and　the　simulated　data　is　4.9%.　The　significance　values　of　normality　test,　homogeneity　test　of　variance,　and　independent　sample　test　are　all　greater　than　0.05,　verifying　that　the　PathFinder　software　can　simulate　passenger　movements　accurately.　Compared　with　the　normal　evacuation　without　optimization　strategies,　the　optimization　degrees　of　the　route　optimization　model　under　three　sets　of　Pareto　solutions　are　16.7%,　15.9%,　and　18.0%,　respectively.　Therefore,　corresponding　evacuation　optimization　strategies　can　be　selected　based　on　specific　evacuation　scenarios,　risk　factors,　service　quality　requirements,　and　other　indicators.　©　2023　Chang＇an　University.　All　rights　reserved.