A novel method based on numerical fitting for oil price trend forecasting. (15th June 2018)

• Record Type: Journal Article
• Authors: Zhao, Lu-Tao; Wang, Yi; Guo, Shi-Qiu; Zeng, Guan-Rong
• Is Part Of: Applied energy. Volume 220(2018)
• Page Start: 154
• Page End: 163
• Publication Date: 2018-06-15
• Subjects: Vector trend forecasting model -- Numerical fitting -- Multi-frequency data -- Crude oil prices; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2018.03.060
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: We innovative definition of time series trends based on numerical fitting methods. The Vector Trend Forecasting Method (VTFM) is developed to oil price forecasting. Exponential and linear functions are better than periodic function in VTFM. VTFM can deduce the trend directly and avoid randomness in oil price forecasting. We forecast mid and long term oil price trends such as monthly, quarterly or semi-annual. Abstract: Crude oil plays an important role in various production processes throughout the world. Changes in oil prices affect economic development, social stability and the residents in a country. Based on a full consideration of the fluctuations in oil prices and discovering the future dynamic trend of oil prices from historical trend features, a vector trend forecasting method that defines the vector trend over a specified length of time and predicts future price trends of crude oil based on the vector trend series of historical crude oil prices is proposed. The core idea behind vector trend forecasting method is to construct the vector trend by using the parameters of a fitting function within a specified interval. Based on the previous linear regression, a variety of non-linear morphological features were selected for numerical fitting, avoiding unity in the price trend and stochastic factors that are difficult to solve in forecast price trends. Combined with an econometric model composed of simultaneous equations, making full use of the characteristic information of the historical vector trend makes the definition of the trend more reasonable and the prediction more accurate. The empirical results show that the percentage error of the fitted real oil price in the vector trend is less than 4%. At the same time, it is found that the numerical fitting result using exponential and quadratic functions are better than that with general linear regression. The forecasting error of the trend is no more than 5%, which is lower than the traditional forecasting accuracy of econometrics and statistical learning models. This study can provide suggestions for oil market investors to understand trends in oil prices and for their investment decision-making, and provide reference for policy makers to stabilize economic markets and people's life.