時間數列模型在原油價格預測之比較分析

A Comparative Analysis of Time Series Models on Forecasting Crude Oil Prices

洪毓婷(YU-TING HUNG)

指導教授 ： 許玉雪

國立臺北大學/商學院/統計學系/碩士(2016年)

摘要

近期原油價格的波動起伏不定，時而增漲時而下跌，然而油價又與我們的生活息息相關，油價的變動往往也牽動著物價的改變，進一步影響著我們日常生活所需，因此事先了解油價的變動趨勢，將有助於相關政策的研擬。據此，本文主要研究目的在於比較分析幾個時間數列模型，試圖找出一個用來預測原油價格的較佳模式，同時進行未來幾年的油價預測。首先，整合過去國內外關於原油價格預測的方法，包含ARMA模式、ARFIMA模式、GARCH模式家族和Markov Switching模式等，利用原油價格的月資料，進行實證分析，比較分析各模式的預測能力，提出一個較佳的預測模式，作為後續預測的基礎。
研究結果發現，配適WTI原油價格的最佳模式為AR(2)- GJR-GARCH(1,1,1)模式，其次依序為具有長期記憶性質的ARFIMA(1,d,0)模式、AR(1)-GJR-GARCH(1,2,1)模式和MS(2)-AR(1)模式；在Brent原油價格所配適的模式中，最佳的預測模式為AR(1)-GARCH(1,1)模式，其次依序為AR(1)-GJR-GARCH(1,2,1)模式、MS(2)-AR(1)模式和 ARFIMA(1,d,0)模式；而Dubai原油價格則是選擇配適具長期記憶性質的ARFIMA(1,d,0)為最佳的預測模式，其次依序為ARMA(2)-GARCH(1,1)模式、AR(2)-GARCH(1,1)模式和MS(2)-AR(1)模式。根據前述各原油價格較佳的配適模式預測未來半年WTI、Brent和Dubai三種原油價格皆有上升的趨勢。

關鍵字

原油價格 ； 預測模式 ； 價格波動 ； 時間數列模式

並列摘要

Recently, the volatility of crude oil prices is fluctuating. The fluctuation of crude oil price would impact many aspects of our daily life. Therefore, it is important to see future trend of the crude oil price. This thesis aims to find a better forecasting model for crude oil prices. Based upon the previous studies, some forecasting models used for crude oil prices are carried out. Time series models including ARMA, ARFIMA, GARCH, and Markov Switching model are used for comparative analysis for crude oil price forcasting. An empirical analysis is conducted based on monthly crude oil prices to see the forecast performance of the forecast models.
The empirical results suggest that AR(2)- GJR-GARCH(1,1,1) model is the best forecasting model for WTI crude oil prices, AR(1)-GARCH(1,1) model is the best forecasting model for Brent crude oil prices, and ARFIMA(1,d,0) model is the best forecasting model for Dubai crude oil prices. The proposed forecasting models are thus used for future forecasting of crude oil prices. It is predicted that crude oil prices would be increased during next six months.

並列關鍵字

crude oil prices ； forescasting model ； price volatility ； time series model

參考文獻

陳盈君（2012），「馬可夫狀態轉換模式與GARCH族群時間數列模式之預測比較分析─玉米期貨價格之實證研究」，國立臺北大學統計研究所碩士論文

王詩惠（2012），「飼料玉米價格之預測–單變量與多變量時間數列模型之比較分析」，國立臺北大學統計研究所碩士論文

Arouri, M.E.H., Nguyen, D.K., and Lahiani, A. (2012), “Forecasting the conditional volatility of oil spot and futures prices with structural breaks and long memory models”, Energy Economics, Vol.34, No.1, pp.283–293.

Balcilar, M., Gupta, R., and Miller,S.M. (2015), “Regime switching model of US crude oil and stock market prices: 1859 to 2013”, Energy Economics,Vol.49 ,pp.317–327.

Baillie, R. T., Bollerslev, T., and Mikkelsen, H. O. (1996), “Fractionally integrated generalized autoregressive conditional heteroskedasticity”, Journal of Econometrics Vol.74, pp.3-30.