Prediction of Daily Precipitation using an Artificial Neural Network Technique combined with Wavelet Decomposition

Rainfall is highly non-linear and complicated phenomena, which require nonlinear mathematical modeling and simulation for accurate prediction. This study comparing the performance of the prediction of one-day-ahead, where Two Feed Forward Neural Network FFNN models were developed and implemented to predict the rainfall on daily for three months (December, January, February). These models are Artificial Neural Network traditional (ANN) model and artificial neural network technique combined with wavelet decomposition (Wavelet- Neural) According to two different methods to build a model using two types of wavelets of Daubechies family (db2, db5). In order to compare the performance of the models in their ability to predict the rains on short-term (for one and two and three-days-ahead) the last months of the period of study, used some statistical standards, These parameters include the Root Mean Square Error RMSE, Coefficient Of Correlation (R).