亞洲貨幣兌美元匯率的線性及非線性時間序列預測模型比較

A comparative study of linear and nonlinear time series forecasting models for the Exchange rate between Asian currency and U.S Dollar

邱瓊儀(Chiung-Yi Chiu)

指導教授 ： 林財川

國立臺北大學/商學院/統計學系/碩士(2016年)

摘要

匯率影響許多經濟決策與外匯市場參與者的行為，若能掌握匯率波動的走勢，有利於企業及投資人決定經營投資策略。本文以1984年1月至2014年6月新加坡幣兌美元、韓幣兌美元、新台幣兌美元之匯率資料，並以新加坡、南韓及台灣每月出口量之資料做為外生變數，以自我迴歸整合移動平均模型、廣義線性自我迴歸模型、類神經網路模型、廣義線性自我迴歸模型及類神經網路模型之混合模型、組合預測模型及非線性組合預測模型等六種模型進行預測，最後以正規均方誤差、方向變化統計量及年報酬率等指標進行預測能力比較。實證發現，新加坡幣、韓幣及台幣在類神經網路模型的預測結果最差，其次是組合預測模型，顯示出新加坡、南韓及台灣的匯率走勢受到了各國每月的出口量影響。

關鍵字

時間序列

並列摘要

Foreign exchange rate will influence lots of economic policy and the behaviors of market players. If we can have a sense of the trend of exchange rate fluctuations, it will be instrumental for us to formulate strategies in dealing with the deliberations resulting from the decision concerns over both corporate finance and individual investment.
In this study, we use the historical statistics of SGD/USD, KRW/USD, USD/NTD exchange rates which were dated from Jan 1984 through June 2014 as base data, coupled with the exogenous variable derived from monthly export volume of Singapore, South Korea and Taiwan respectively, then run the forecast by utilizing Autoregressive Integrated Moving Average Model, Generalized Linear Auto-Regression Model, Artificial Neural Network Model, The Hybrid Methodology Integrated GLAR with ANN, The Combined Forecasting Model, The Nonlinear Ensemble Forecasting Model separately. In the end, we further utilize normalized mean squared error, directional change statistics and annualized return as indicators to make a comparison analysis on the forecast capability.
Empirical results illustrate that SGD/USD, KRW/USD, USD/NTD exchange rate utilized Artificial Neural Network Model and The Combined Forecasting Model reveals a worse forecast performance. The conclusion demonstrates that the trend of foreign exchange rate of Singapore, South Korea and Taiwan were influenced by their own monthly export volume. Judging from three indicators, we can conclude that the forecast result derived from directional change statistics and annualized return demonstrate a similar outcome.

並列關鍵字

time series

參考文獻

Akaike, Hirotugu. (1974). A new look at the statistical model identification. Automatic Control, IEEE Transactions on, 19(6), 716-723.

Bates, John M., and Granger, Clive W.J. (1969). The combination of forecasts. Or, 451-468.

Box, George E. (1976). P, and Jenkins, GM,“Time Series Analysis: Forecasting and Control,”. Time Series and Digital Processing.

Della Corte, Pasquale, and Tsiakas, Ilias. (2012). Statistical and economic methods for evaluating exchange rate predictability. Handbook of exchange rates, 221-263.

Denton, James W. (1995). How good are neural networks for causal forecasting? The Journal of Business Forecasting, 14(2), 17.