Monthly Stream Flow Prediction Using Support Vector Machine Based on Principal Component Analysis

The main goal of this research is to evaluate the role of input selection by Principal Component Analysis (PCA) on Support Vector Machine (SVM) performance for monthly stream flow prediction. For this purpose, SVM is used to predict monthly flow as a function of 18 input variables. PCA is subsequently employed to reduce the number of input variables from 18 to 5 PCs which are finally fed into the SVM model. SVM and PCA-SVM models are evaluated in terms of their performance using a developed statistic by the authors. Findings show that preprocessing of input variables by PCA improved SVM performance.