Prediction of Rainfall Time Series Using the Hybrid DWT-SVR-Prophet Model

Accurate rainfall prediction remains a challenging problem because of the high volatility and complicated essence atmospheric data. This study proposed hybrid model (DSP) that combines advantages discrete wavelet transform (DWT), support vector regression (SVR), Prophet to forecast First, time series is decomposed into high-frequency low-frequency subseries using (DWT). The SVR models are then used predict subsequences, respectively. Finally, predicted determined by summing values each subsequence. A case in China conducted from 1 January 2014 30 June 2016. results show DSP provides excellent prediction, with RMSE, MAE, R2 6.17, 3.3, 0.75, yields higher accuracy than three baseline considered, ranking as follows: > SSP SVR. In addition, quite stable can achieve good when applied data various climate types, RMSEs ranging 1.24 7.31, MAEs 0.52 6.14, 0.62 0.75. may provide novel approach for forecasting readily adaptable other predictions.