Based　on　kalman　filtering　theory,　a　improved　Kalman　filter　short-term　prediction　model　is　put　forward　and　the　solving　process　is　presented　after　the　characteristic　analysis　of　rail　transit　platform.　The　data　acquisition　and　example　analysis　are　carried　out　on　the　island　platform,　side　platform,　common　platform　and　transfer　platform　with　large　passenger　flow　and　obvious　change　of　passenger　flow　in　Beijing.　The　results　show　that　the　average　absolute　error　of　the　model　is　0.299,　the　mean　square　error　is　34.094,　and　the　equal　coefficient　is　0.923,　which　reveals　that　the　proposed　model　can　effectively　predict　the　short-term　subway　passenger　flow.　Compared　with　the　traditional　Kalman　filtering　prediction　method,　the　improved　Kalman　filter　short-term　passenger　flow　forecasting　method　can　improve　the　real-time　information　of　prediction,　reduce　the　average　absolute　error　by　0.448,　and　has　higher　prediction　accuracy.　©　2017,　Editorial　Department　of　Journal　of　Wuhan　University　of　Technology.　All　right　reserved.