國際黃金價格預測─ARIMA模式與迴歸模式之比較(民國87-99年)

The comparison of ARIMA model and regression model on the forecast of gold price；1998-2010

李佳倩(Chia-Chien Lee)

指導教授 ： 林逾先

國立臺北科技大學/管理學院/商業自動化與管理研究所/碩士(2011年)

https://doi.org/10.6841/NTUT.2011.00515

摘要

黃金其成熟且歷史悠久的特性，一直被金融市場視為熱門的投資工具，但如何預測金價的漲跌走勢，一直是許多學者想要探討的問題。若能得知影響黃金價格的關鍵因素，即可建構一套相對報酬較佳之投資方法。

本研究採ARIMA與迴歸模式進行相關研究，並針對此二模式予以比較。資料採用1998年1月至2010年7月期間之黃金價格、美元匯率、白銀價格、消費者物價指數、七大工業國失業率、道瓊工業指數等經濟指標，透過最佳分組迴歸建立多元迴歸模式；並以同期間之國際黃金價格為時間數列資料建立ARIMA模式。

最後本研究分別對2010年8月到9月間國際黃金價格進行預測，並將其結果與實際值進行檢測，評估其預測效力。

研究結果發現，迴歸模式顯示國際黃金價格變化受到白銀價格及美元匯率二者的影響較為顯著。另外，本研究亦發現ARIMA (0,2,1)可作為預測黃金價格變化的合適ARIMA模式。用判定係數(R2)及均方誤差(MSE)兩種指標比較迴歸模式與ARIMA(0,2,1)模式，我們發現其ARIMA(0,2,1)的R2較大且MSE較小，可獲得較佳黃金價格的預測效力。

關鍵字

預測 ； 黃金價格 ； 迴歸模型 ； ARIMA

並列摘要

Many investors prefer the gold market due to its history and maturity. However, understanding how to make correct investments in a fluctuating gold market would be the main challenge when investing in the gold market. If one can understand the factors that affect the gold price, one would be able to make better forecast on the future gold price through an appropriate model.
This research adopts ARIMA model and regression model towards the forecast of gold price using the data from January, 1998 to July, 2010 that includes sliver price, US dollar exchange rate, consumer price index, Dow Jones Industrial Average, unemployment rate of G7 countries. Both regression model and ARIMA model are identified through Best Subsets Regress, and time series patterns, respectively.
The research result from regression model indicates that gold price is affected significantly by silver price and US dollar exchange rate. In addition, it finds that ARIMA(0,2,1) is an appropriate ARIMA model in the prediction of gold price. Using the R-square and Mean Square Error (MSE) as indicators, we find ARIMA (0,2,1) model has higher R-square value and lower MSE, making it a better model in the prediction of gold price.
Prediction on the period of August-September, 2010 are conducted and their results compared with the actual values for August-September, 2010 to assess the predictability of the model.

並列關鍵字

Predict ； Gold price ； Regression ； ARIMA

參考文獻

[4] 李振宇、陳至安和簡郁紘，應用線性迴歸模型，台北市：美商麥格羅•希爾國際股份有限公司 台灣分公司，2005。

[12] B. Sharma, Edward Y. H. Lin and E. Otuteye, " An ARIMA model for Canadian union membership growth," Applied Economics, 1911-1985, vol. 24, 1992, pp. 1035-1041.

[14] C. Eugene, P. C. Gauri and M. T. Alan, " MEASURING MARKET INTEGRATION : FOREIGN EXCHANGE ARBITRAGE AND THE GOLD STANDARD, 1879-1913," The Review of Economics and Statistic, vol. 86, no.4, 2004.

[15] C. Liu, Price Forecast For Gold Futures Based on GA-BP Neural Network, Master. Thesis, University of Shanghai, China, 2009.

[17] E. A. Selvanathan, "A Note on the Accuracy of Business Economists’ Gold Price Forecasts ," Australian Journal of Management, vol. 16, no.1, 1991, pp. 91-95.