Passenger　flow　predictions　are　of　great　significance　to　bus　scheduling　and　route　optimization.　In　this　paper,　a　novel　algorithm,　namely　the　Spatial-Temporal　Graph　Sequence　with　Attention　Network　(STGSAN),　was　proposed　to　predict　transit　passenger　flow.　The　algorithm　mainly　focused　on　the　following　three　aspects:　(1)　a　graph　attention　network　(GAT)　was　used　to　capture　the　spatial　correlation　of　various　bus　stops;　(2)　to　make　full　use　of　the　historical　and　real-time　data,　a　bidirectional　long　short-term　memory　and　attention　mechanism　was　conducted　to　extract　the　temporal　correlation　of　historical　ridership　at　bus　stations;　and　(3)　external　factors　that　affect　passenger　choices　were　taken　into　account.　We　conducted　an　experiment　using　field　data　collected　in　Urumqi,　China.　After　comparison　with　five　other　models,　the　proposed　model　was　proven　to　have　excellent　performance　prediction.