Short　term　traffic　prediction　is　a　common　Spatial-Temporal　sequence　training　and　learning　prediction.　This　paper　presents　a　new　prediction　model　STWLSTM.　Firstly,　the　model　is　decomposed　and　reconstructed　by　wavelet　analysis,　and　then　LSTM　is　trained.　The　LSTM　prediction　results　are　decomposed　by　wavelet,　and　the　wavelet　high-frequency　coefficients　of　historical　time　data　and　first-order　neighborhood　spatial　data　are　fused　before　reconstruction　to　obtain　the　final　results.　The　experimental　results　show　that　the　prediction　results　of　STWLSTM　model　conform　to　the　inherent　attributes　of　traffic　prediction:　periodic　stability　and　mutation.　　©　2022　IEEE.