In　the　process　of　high　altitude　detection,　due　to　the　hysteresis,　solar　radiation　and　other　factors,　there　is　a　deviation　between　the　measured　data　of　temperature　sensor　and　the　standard　data.　To　solve　this　problem,　we　combine　depth　neural　network,　wavelet　function,　SVM　and　XGBoost　to　propose　an　error　prediction　model.　Morlet　wavelet　is　used　as　the　activation　function　of　neural　network　to　improve　the　prediction　ability.　The　stacking　integrated　learning　method　is　used　to　build　a　cascade　prediction　model　to　achieve　extreme　gradient　promotion.　By　collecting　the　real　data　of　meteorological　observation,　the　dataset　is　established,　and　the　proposed　method　is　evaluated　on　this　dataset.　The　experimental　results　show　that　compared　with　the　traditional　model,　the　improved　model　has　certain　effectiveness,　MSE　reduces　0.173,　effectively　overcomes　the　influence　of　solar　radiation,　and　improves　the　measurement　accuracy　of　the　sensor.　Moreover,　this　method　has　strong　generalization　and　can　be　easily　extended　to　other　data　prediction　and　regression　tasks.　©　2020　IEEE.