Comparison of discrete and continuous wavelet-multilayer perceptron methods for daily precipitation prediction

Wavelet transforms are combined with predictive methods to develop prediction approaches so that the prediction accuracy can be improved in hydrologic predictions. Although the wavelet transform generates several subseries that show similar characteristics, the predictive method is used to develop the model using those subseries. There are several examples of these kinds of combined models, such as wavelet-multilayer perceptron (MP), wavelet fuzzy, wavelet autoregressive, and so forth. Generally, discrete wavelet transformation is used in combined models rather than continuous wavelet transform for unexplained reasons. As a result, in this study emphasis was placed on the comparison of the continuous wavelet-multilayer perceptron (CWT-MP) and discrete wavelet-multilayer perceptron (DWT-MP) models, which were also compared with the stand-alone MP model. Daily precipitation time series from two stations were used in the model development and comparison process. The current precipitation values were predicted from previous precipitation values. Various scenarios were used for the establishment of the models. Mean square error (MSE), coefficient of efficiency (CE), and score skill (SS) were used as model evaluation criteria, and it was observed that the prediction performance of MP was significantly improved by using wavelet transforms as preprocessing techniques. However, the CWT-MP models were found to be better than the DWT-MP models based on the results of the evaluation criteria.