Speed　is　an　important　factor　for　traffic　safety　evolution.　The　use　of　ITS　technology　in　speed　management　with　real-time　information　of　urban　freeway　is　one　of　strategies　to　enhance　road　safety.　This　paper　presents　methods　for　prediction　short-term　traffic　flow　speed　on　Beijing　urban　freeway　with　real　time　information　from　inductive　loops.　The　source　data　sets　including　traffic　volumes,speed　and　occupancy　which　are　collected　24h/day　over　several　years　on　Beijing　ring　road.　Traffic　flow　is　divided　into　three　statuses,　including　free　flow,　transition　status　and　congestion　according　to　occupancy.　To　achieve　objective　prediction　results,　wavelet　technology　is　applied　to　de-noising　process.　The　artificial　neural　network,　which　does　not　require　any　pre-defined　underlying　relationship　between　dependent　and　independent　variables,　is　a　powerful　tool　in　dealing　with　prediction　problems.　In　this　paper,　RBF　network　is　designed　for　predicting　speed　for　future　five　minutes.　Results　show　that　the　proposed　RBF　network　model　produces　reliable　estimates　of　vehicle　speed　for　three　various　traffic　conditions,　especially　congestion　condition.