Elham Hajikaram; Roya Darabi
Abstract
Anticipating process of crude oil prices and its fluctuations volatility has always been one of the challenges the traders face in the exchange oil markets. This study estimates the Brent crude oil daily price forecast with a proposed hybrid model. The sample is Brent North Sea crude oil daily prices ...
Read More
Anticipating process of crude oil prices and its fluctuations volatility has always been one of the challenges the traders face in the exchange oil markets. This study estimates the Brent crude oil daily price forecast with a proposed hybrid model. The sample is Brent North Sea crude oil daily prices from July 2008 to July 2016 that is selected from the total oil daily prices in all of the oil markets. In this research, a model for combining statistical and artificial intelligence (PCA-SVR) methods is presented. With regard to the superiority of the accuracy of the prediction of the support vector regression model (SVR) in comparison with other predictive methods in past studies, the main goal in this research is to improve the prediction of the support vector regression using the initial pre-processing of data by principal components analysis (PCA). To do research, after carrying out a static test, using principal components analysis, the input variables are converted into the principal components that cover the entire data scattering and considered as an input to the prediction model. Then, using supporting vector regression model and simulate it in MATLAB software we predict daily price of Brent crude oil. In order to compare the performance of the SVR and PCA-SVR models, we used the paired comparison test. The result of this study was that the initial pre-processing by means of the principal components analysis on the data gives rise to reducing suggested model error
