Using principal component analysis to inputs the effective rainfall estimates based on entries to help support vector machine and artificial neural network

Atmospheric precipitation management and efficient use of these resources are extremely helpful to the management of water resources and it has the basic role in the management of the water resources of hydrological parameter estimate as well. In this research, rainfall estimate of three synoptic station situated in Astara, Lahijan and Jirandeh in Gilan province has been done using artificial neural network (ANN) and support vector machines (SVM). The principal component analysis method (PCA) was applied to determine the data pre-processing and input data. According to the results, the main component analysis method for synoptic stations as well as for the five main components of Astara and jirandeh, and for the four main components of Lahijan station is selected. The results of model-making indicate that the artificial neural network model based on principal component analysis (PCA-ANN) in Astara and Jirandeh stations in order with average 2.74 mm squares of 2.62and mm backup-based support vector machine model analysis of the main components (PCA-SVM) mean squares of squares in Lahijan station with error 2.53 mm can be selected as selected models for the aforementioned stations. Finally, with respect to the results can be fitted such that the methods used for data pre-processing in this research to predicate the rainfall as well as the SVM in Lahijan station and model Ann model in Astara and Jirandeh stations have been acceptable.