Improved Crude Oil Price Forecasting With Statistical Learning Methods

Abstract

Reliable forecasts of the price of oil are of interest for a wide range of applications. For example, central banks and private sector forecasters view the price of oil as one of the key variables in generating macroeconomic projections and in assessing macroeconomic risks. Of particular interest is the question of the extent to which the price of oil is helpful in predicting recessions. This paper presents a statistical learning method (SLM) based on combined fuzzy system (FS), artificial neural network (ANN), and support vector regression (SVR) to cope with optimum long-term oil price forecasting in noisy, uncertain, and complex environments. A number of quantitative factors were discovered from this model and used as the input. For verification and testing, the West Texas Intermediate (WTI) crude oil spot price is used to test the effectiveness of the proposed learning methodology.Empirical results reveal that the proposed SLM-based forecasting can model the nonlinear relationship between the input variables and price very well. Furthermore, in-sample and out-of-sample prediction performance also demonstrates that the proposed SLM model can produce more accurate prediction results than other nonlinear models.Keywords: crude oil price, fuzzy system (FS), artificial neural networks (ANNs), support vector regression (SVR) IntroductionCrude oil has been playing an increasingly important role in the world economy since nearly two thirds of the world's energy demands are met from crude oil (Alvarez-Ramirez, Soriano, Cisneros, & Suarez, 2003). It is said that crude oil is also the world's largest and most actively traded commodity, accounting for over 10% of total world trade (Verleger, 1993). Like most commodities, crude oil price is mainly influenced by international politics, economy, military affairs, diplomacy, and other factors. The frequent change of these factors makes oil price show uncertainty, mutagenicity, and randomness. In addition, crude oil products are one of the world's major commodities with a high volatility level. They are traded in New York Mercantile Exchange (NYMEX) market together with other energy and mineral commodities. Simultaneously, the volatility of this crude oil price depends on demand and supply of the commodity, level of inventories, and economic indicators.In the past decades, traditional statistical and econometric techniques, such as linear regression (LinR), co-integration analysis, Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models, naive (2001) suggested a semi-parametric statistical method for short-term oil price forecasting based on the GARCH properties of crude oil price. Similarly, Barone-Adesi, Bourgoin, and Giannopoulos (1986) suggested a semi-parametric approach for oil price forecasting. Gulen (1998) used a co-integration analysis to predict the West Texas Intermediate (WTI) price.Ye, Zyren, and Shore (2006) presented a simple econometric model of WTI prices, using OECD petroleum inventory levels, relative inventories, and high-and low-invento...

An Auxiliary Index for Reducing Brent Crude Investment Risk—Evaluating the Price Relationships between Brent Crude and Commodities

“…Yu et al (2008) made a forecast on crude oil prices with an EMD-based neural network [19]. Slim (2015) improved the forecasting of crude oil prices with statistical learning methods (SLM) based on a combined fuzzy system (FS), artificial neural network (ANN), and support vector regression (SVR) [20]. He et al (2016) showed a significant performance improvement in forecasting WTI and Brent Crude oil prices using a multivariate EMD-based model [21].…”

Hybridization of Support Vector Regression and Arima Models with Genetic Algorithms for Predicting Crude Oil Price

“…Some nonlinear and emerging machine learning models, such as Support Vector Machines (SVMs), Random Forest (RF), Artificial Neural Networks (ANNs), and Genetic Programming (GP), offer effective solutions to nonlinear crude oil price prediction because they overcome the shortcomings of conventional statistical and econometric models [11]. To model and forecast the price of crude oil, for instance, Slim [12] proposed a new Statistical Learning Technique (SLM) based on a combined Fuzzy System (FS), Support Vector Regression (SVR), and ANN. Similarly, several studies revealed that models based on machine learning frequently had some advantages over statistical-based models [13][14][15][16].…”