Discharge prediction in phetchaburi basin using a combination of wavelet and cross correlation

Discharge prediction is an essential component in water management systems. To obtain an accurate prediction model, we need a good preprocessing method for extracting actually important features of the discharge data. Thus, we propose a new combinational method which integrates Correlation Coefficient Analysis and Wavelet Decomposition. The processed discharge data from both methods are then used as input for two classification methods, namely Backpropagation Neural Networks and Multiple Linear Regression. In our experiment, we tested our method based on the real world data from the Phetchaburi river basin, Thailand. The obtained model achieved lower error rate than ones from other existing methods.