In this study, a coupled SVM-CEEMDAN-BiLSTM (SCB) model was proposed to improve the accuracy of daily precipitation forecasting. The support vector machine (SVM) method was employed to classify the complex meteorological data at first. Next, the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) was applied to decompose the precipitation series into some subseries. Then, the Bi-directional Long Short-Term Memory (BiLSTM) was constructed to predict the subseries. All the predicted subseries were summed and transformed as the daily precipitation prediction. Finally, the predicted accuracy was examined by four evaluation indexes for 14 stations over the Poyang Lake Basin. Results show that the prediction of the SCB model is more skillful in the southern region than in the northern region of the basin. Moreover, the SCB model outperforms other models at all stations. The SCB model improves daily precipitation prediction accuracy and provides the best prediction performance compared with other models.