- 學位論文
整合搜尋趨勢及建模誤差篩選預測國際原油價格
Forecasting International Oil Price with the Integration of Google Trends and Model Residual Error Filtering
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
油價對於經濟發展有著連動關係，原油價格大幅波動對於石油進口國和石油出口國有重大的影響，而從油價歷史來看，油價波動的不確定性通常是因為突發性事件所造成，例如:金價、政治因素、股市、美元匯率、投資因素、OPEC產出、美國制裁好戰國家、氣候因素及石油庫存等因素。這些事件引來了學術界對於油價波動的原因與石油價格波動對總體經濟的影響，以及作為主要石油輸出國組織(OPEC)在國際石油市場上的地位和作用等等這些問題的更多更深入的探討，本研究期望對於從事原油相關產業或相關學術研究者，提供預測油價的相關研究。 本研究以時間序列灰色預測以及比較模型線性迴歸、M5模式樹、隨機森林模型預測西德州及布蘭特原油價格。使用變數為OPEC原油產量、OPEC剩餘原油產能、OECD石油消費量、OECD石油庫存量、美元指數、美國原油淨進口、Kilian經濟指數、原油期貨價格。利用灰色建模誤差、搜尋趨勢及整合策略的篩選策略組合基礎模型。採用MAPE、RMSE、MAE評估標準以及單變量檢定、Duncan事後檢定來評估預測模型的準確度，結果顯示利用搜尋趨勢篩選策略將灰色預測與M5模式樹組合之模型相對於其他模型有較佳的準確度。
並列摘要
There's a close relation between oil price and economic development. When crude oil prices increase or drop dramatically, that will have a great impact on oil-importing and oil-exporting countries. Throughout the history of oil prices, the uncertainty of oil prices changes are due to sudden events such as gold prices, politics, stock markets USD exchange rate, investments, OPEC output, USA fighting against warlike countries, climate, and oil in stock factors. These events cause the academia have further discussion about these factors. First, they need to know why the events lead to oil prices change, which also influence the economy of the world. Besides, they need to discuss that, as a member of OPEC, what important place they occupy and what functions they have. The study expect to offer basic information of oil prices change to those have concern over crude oil-related industries. The study uses Grey Prediction and Linear Regression, M5 Model Trees, Random Forests for forecasting West Texas oil price and Brent oil price. Crude oil variables includes OPEC oil Production, OPEC Spare Crude Oil Production Capacity, OECD Petroleum Consumption, OECD Petroleum Stocks, USA Dollar Index, USA Net Imports of Crude Oil, Kilian’s Global economic index, Crude oil futures prices. The study also apply Grey Model Residual Error, Google trends and integration Strategy to form combined model. And then it adopts MAPE, RMSE, MAE, analysis of variance and Duncan multiple range test to assess model. It turns out that by using Google Trends strategy, the combination of Grey Prediction and M5 model trees has better accuracy.
參考文獻
7.黃哲賢(2012)。應用線性迴歸模型於射頻IC 特性預測。未出版碩士論文，國立交通大學，新竹。
1.Archer, J. & Kimes, R. (2008). Empirical characterization of random forest variable importance measures, Computational Statistics & Data Analysis,52, 2249 - 2260.
2.Ahmad Kazem, Ebrahim Sharifi, Farookh Khadeer Hussain, Morteza Saberi & Omar Khadeer Hussain, (2012). Support vector regression with chaos-based firefly algorithm for stock market price forecasting,Applied Soft Computing,(24),1-25.
3.Bates, J. M. & Granger C.W.J.(1969).The Combination of Forecasts, Operational Research Quarterly,(4),451-468.
7.Hotelling,H. (1931).The Economics of Exhaustible Resources.The Journal of Political Economy,39(2),137-175.