With　the　development　of　Information　and　Communication　Technology　(ICT),　the　service　provided　by　cloud　data　centers　has　become　a　new　pattern　of　Internet　services.　The　prediction　of　the　number　of　arriving　tasks　plays　a　crucial　role　in　resource　allocation　and　optimization　for　cloud　data　center　providers.　This　work　proposes　a　hybrid　method　that　combines　wavelet　decomposition　and　autoregressive　integrated　moving　average　(ARIMA)　to　predict　it　at　the　next　time　interval.　In　this　approach,　the　task　time　series　is　smoothed　by　Savitzky-Golay　filtering,　and　then　the　smoothed　time　series　is　decomposed　into　multiple　components　via　wavelet　decomposition.　An　ARIMA　model　is　established　for　the　statistical　characteristics　of　the　trend　and　components,　respectively.　Finally,　their　prediction　results　are　reconstructed　via　wavelet　reduction　and　the　predicted　number　of　arriving　tasks　is　obtained.　Experimental　results　demonstrate　that　the　hybrid　method　achieves　better　prediction　results　compared　with　some　typical　prediction　methods　including　ARIMA　and　neural　networks.　©　2018　IEEE.