摘要
自从1973年西方各主要工业国家实行浮动汇率制度以来,外汇市场上的汇率波动频繁,不论是参加外汇交易的个人还是银行、企业乃至于主权国家,都面临着巨大的外汇风险。这在近二三十年来频繁发生的货币危机中得到了体现,对危机发生过的经济发展都造成了相当程度的负面影响。同时,外汇市场的交易规模已经远远超过股票、期货等其他金融商品市场,成为当今全球最大的金融市场。外汇市场上的参与者,比如进口商、出口商、投资组合的管理者以及各国中央银行等,如同生态系统中的各级食物链,市场规模的快速增大,让这种食物链关系更加复杂。这使得通过各种方法预测汇率的波动,变得极为迫切。对于汇率波动的预测不仅可以帮助外汇市场上的投机者制定交易策略,降低风险,更主要的是能够使得各主权国家可以根据市场的情况对可能的金融危机进行预警,防范金融风险,维护国家金融安全。
本文首先总结归纳了影响汇率长期波动和短期波动的主要因素,现有的汇率预测方法研究主要是从基础变量和技术角度两个方面进行的,通过对这些工作进行比较,进而给出了基本的评价。但是由于外汇市场交易的复杂性,传统的宏观汇率决定模型已经难以实际预测汇率的短期波动。正是在这一背景之上,本文选择技术方法对短期的汇率波动进行预测,选取欧元美元的小时收盘价作为预测对象。
本文的重点在于提出了基于市际(intermarket)信息的欧元对美元的非线性预测模型。我们使用多层前馈神经网络(MFNN-Multilayer Feedforward Neural Networks)作为非线形预测模型的计算形式。论文详细阐述了MFNN的原理,尤其是采用后向传递(BP)训练算法的MFNN的计算模型,并对现有的基于神经网络的汇率预测方法进行了大致的回顾和评价。接下来通过构建单市场和多市场的时间序列模型,对欧元美元的小时收盘价进行预测,并通过方向正确的命中率的方法来检验实际的预测效果。在单市场模型中,以欧元美元小时收盘价的历史数据作为神经网络的输入,实证表明,单市场模型的预测效果并不十分理想。在多市场模型中,首先分别将欧元美元、英镑美元、美元瑞士法郎、美元日元等相关货币对的历史价格信息作为输入,得到较为理想预测结果,其方向正确的命中率达到62.63%。多市场模型中还将美元指数以及经过换算后的欧元指数、英镑指数、瑞士法郎指数和日元指数作为输入进行预测,但实证表明,与单市场模型相比,其预测效果并没有明显提高。
Since the major Western industrial countries put the floating exchange rate system into practice in 1973, exchange rate on the foreign exchange market fluctuates frequently. Everyone who participates in foreign exchange transactions are facing with a huge foreign exchange risk,no matter individuals or banks, businesses, and even sovereign states. The risk has been reflected by frequent occurrence of currency crises in the two or three decades ,which led to negative impact on the economics.Meanwhile, the scale of the foreign exchange market trading has been far more than the stock、futures and other financial products markets, become the largest financial market. The foreign exchange market participants, such as importers, exporters, portfolio managers and central banks, etc., just like the food chain at all levels in the ecosystem.As the rapid increase of market size, the food chain relationships become more complex.This makes the forecasting of exchange rate through various methods has become extremely urgent. This will not only help the speculators on the foreign exchange market make trading strategies and reduce risk, but also can make the sovereign state carry early warning for possible financial crisis ,and keep the financial risk away, safeguard national financial security.
This paper summarized the short-term and long-term factors which impact the exchange rate fluctuations at first. Prediction of the current exchange rate is mainly based on basic variables and technical point of view, this paper give a simple evaluation on such work. However, due to the complexity of the foreign exchange market transactions, the traditional macro exchange rate determination models have been difficult to predict short-term exchange rate fluctuations.On this basis, the research chose the technology method to predict the short-term exchange rate, select the hourly closing price of the EURUSD as forecast object.
The focus of this paper is it proposed the EURUSD non-linear prediction model based on intermarket information. We use a multilayer feedforward neural network(MFNN) prediction model for the calculation of non-linear form. This paper elaborate the principle of MFNN in detail, especially the BP training algorithm calculation model MFNN the BP training algorithm.In the second part of this article, the theory of artificial neural networks is elaborated, especially the content of BP neural network, and also,the exchange rate forecasts based on neural network methods were generally reviewed and evaluated. Next, by constructing a single variable and multivariate time series model, predict the hourly closing price of EURUSD, And test the actual prediction effect by correct direction hit rate.
In the single variable model, make the historical hourly closing price of EURUSD as input data, the research shows that the predicted effect is not very satisfactory. In the multivariable model, the historical hourly closing price of relevant currency like EURUSD、GBPUSD、USDCHF、USDJPY are selected as input data, the correct direction hit rate can reach 62.63%. In the research,The USD index and other converted currencies index such as EUR index、GBP index、CHF index、JPY index are also selected as input data to forecast the target exchange rate, there is no significant improve compared with single variable model.
引文
[1]凯茜·莲恩.外汇市场即日交易[M].广州:广东经济出版社,2007:51-70
[2] Osborne M.Brownian Motion in Stock Market[J].Operations Research,1959,7(6): 807-811
[3] Osborne M.The Random Character of Stock Market Price[M].Cambridge:MIT Press, 1964
[4] Muller U,Dacorogna M,Dave R.,Olsen R..Volatilities of different time resolutions- analyzing the dynamics of market components[J].Journal of Empirical Finance.1997 ,4(23):213-239
[5] Peters E E.Chaos and Order in the Capital Markets:A New View of Cycles,Price,and Market Volatility[M].New York:John Wiley and Sons,1991
[6] Peters E E.Fractal Market Analysis:Applying Chaos Theory to Investment & Economics[M]. New York:John Wiley and Sons,1994
[7] Pan H P, Sornette D. and Kortanek K.Intelligent Finance-An introduction [J]. China Journal of Finance. 2005, 3(3): 99-106
[8] Pan H P, Sornette D. and Kortanek K.Intelligent Finance-An Emerging Direction [J]. Journal of Quantitative Finance, 2006, 6(4): 273-277
[9] R.meese,k.rogoff.Empirical Exchange Rate Models of the Seventies:Do They Fit Out-of-sample? Journal of International Economics,1983,(14):3—24
[10] R.meese,A.Rose.An Empirical Assessment of Non-linearities in Models of Exchange Rate Determination.Review of Econometrics Studies,1991,(51):601-609
[11] R.MacDonald.M.Taylor.Exchange Rate Economics:a Survey .IMF Staff Papers,1992, (39):1-57
[12]张忠杰.ARIMA模型在汇率预测中的应用.经济理论研究,2005,(7):39-4l
[13]惠晓峰,柳鸿生,胡伟等.基于时间序列GARCH模型的人民币汇率预测.金融研究,2003,(5):99-105
[14] Z Ding,C W J Granger,R F Engle.A long memory property of stock market returns and a new model.Joumal of Empirical Finance,1993,(1):349—372
[15] Z Ding,C W J Granger.Modeling volatility persistence of speculative returns:a new approach.Journal of Econometrics,l996,73(1):l85-215
[16]陈健,冉茂盛,钱灿.人民币实际有效汇率波动的特征.统计决策,2005,(10s):79-80
[17] L Kilian,MP Taylor.Why is it so difficult to beat the random、walk forecast of exchange rates?.Joumal of International Economics,2003,60(1):85-107
[18] N Sarantis.On the short-term predictability of exchange rates:A BVAR time- Varying parameters approach.Journal of Banking&Finallce,2006,30(8):2257-2279
[19]黄松.Bayes网络在汇率趋势预测中的应用.计算机与现代化,2006,(3):9-13
[20] Charles Engel.Can the Markov Switching model forecast exchange rates?.Journal of International Economics,1994,36(1):151-165
[21]任敬喜,王文哲.基于马尔可夫链的欧元汇率走势分析.统计与决策,2005,(04X):118- 119
[22] Claudia La、vrenz,Frank westerhoff Explaining exchange rate Volatility with a Genetic Algorithm.International Conferences of the Society for Computational Economics on Computing in Economics and Finance,2000,5(1):11-13
[23] A N Refenes,M AzemaBavac,L Karoussos.Currency exchange rate prediction and neural network design strategies.Neural Computing & Applications,1993,1(1): 46-58
[24] Mark Leung,AnSing Chen,Hazem Daouk.Forecasting exchange rates using general regression neural networks.Computers & 0perations Research,2000,27(11-12):1093- 1110
[25]何红弟,高亮,王文凯,等.基于遗传神经网络的汇率价格短期预测[J].上海大学学报(自然科学版).2005(2):103-106
[26] De Grauwe P.,Dewachter H.Chaos in the Dornbusch model:the role of fundamentalists and chartists.Open Economies Review,1993,4(4):351—379
[27] Mandelbrot B.B.The variation of certain speculative prices.Journal of Business,1963,36(4):394-419
[28] Bera A.K.A test for conditional heteroskedasticity in time series models.Journal of Time Series Analysis,1992,13(6):501—519
[29] Lyons R.The microstructure approach to exchange rate.USA:MIT Press,2001,112- 135
[30] Zakoian J.M.Threshold heteroskedastic models.Paris:Manuscript CREST INSEE, 1990
[31] Glosten L.R.,Jagannathan R.,Runkle D.E.On the relation between the expectedvalue and volatility of the nominal excess return on stocks.Journal of Finance, 1993,48(5):1779-1799
[32] Brock W.A.,Hsieh D.,LeBaron B.Nonlinear dynamics,chaos and instability:theory and evidence.Cambridge,Mass:MIT Press,1991,47-56
[33] Dacorogna M.M.,Muller U.A.,Nagler R.J.,et a1.A geographical model for the daily and weekly seasonal volatility in the foreign exchange market.Journal of International Money and Finance,1993,12(4):413-438
[34] Bollerslev T.A conditionally heteroskedastic time series model for speculative prices and rates of return,The Review of Economics and Statistics,1987,69(3):542 -547
[35] Granger C.,Joyeus W.Long memory relationships and the aggregation of dynamic models.Journal of Econometrics,1980,14:227-238
[36] Hosking J.Fractional differencing.Biometrika,1981,68(1):165-176
[37]庄新田,黄小原.股价指数的自相关与标度不变性分析.东北大学学报(自然科学版), 2002,23(6):542-545
[38]奚媛媛.人民币汇率与主要非美元货币汇率的相关性分析.南京审计学院学报,2007, 4(1):34-37
[39]张良均、曹晶、蒋世种.神经网络实用教程[M].北京:机械工业出版社,2008
[40]谢赤,欧阳亮.汇率预测的神经网络方法及其比较[J].财经科学2008(5):47-53
[41] Refenes. Constructive learning and its application to currency excha nge rate forecasting. In: Neural networks in finance and investing: using artifi cial intelligence to improve real world performance, 1993, 465-493
[42] De Matos. Neural networks for forecasting exchange rates: [disserta tion n]. Canada: The University of Manitoba, 1994
[43]魏巍贤,蒋正华.汇率的神经网络预报模型及其实例分析.预测,1995,14(2):67-69
[44] Gioqinang Zhang, Michael Y. Hu. Neural Network Forecasting of the Br itish Pound/US Dollar Exchange Rate[J]. Omega, Int. J. Mgmt Sci, 1998, 26(4):495-506
[45]惠晓峰,胡运权,胡伟.基于遗传算法的BP神经网络在汇率预测中的应用研究,数量经济技术经济研究,2002,2,80-83
[46]欧阳亮.基于小波分析与神经网络的汇率组合预测研究:[硕士学位论文],长沙:湖南大学.2008
[47] Mona R. El Shazly, Hassan E. El Shazly. Comparing the forecasting pe rformanceof neural networks and forward exchange rates[J]. Journal of Multina tional Financial Management, 1997(7): 345-356
[48]杨炘,马洪波.人工神经网络在中长期汇率预侧中的应用[J].系统工程,1999,17(1): 18-24
[49] HUI Xiao-feng,LI Zhe,WEl Qing-quan.Using fuzzy neural networks for RMB/USD real exchange rate forecasting[J]. Journal of Harbin Institute of Te chnology (New Series), 2005, 12(2): 189-192
[50]詹亮.基于神经网络的外汇市场预测.[硕士学位论文],武汉:华中师范大学
[51]陈莹.美国股指SP500的走势预测模型——基于移动平均线、BP神经网络和小波神经网络的实证研究.[硕士学位论文],成都:电子科技大学
[52] Bates JM, Granger CWJ. The combination of forecasts[J]. Operation s Research Quarterly, 1969, 20: 451-68
[53] Michael Y. Hu, Christos Tsoukalas. Combining conditional volatilityforecasts using neural networks: an application to the EMS exchange rates[J].Journal of International Financial Markets, Institutions and Money, 1999, (9):407-422
[54] Fang-Mei Tseng, Hsiao-Cheng Yub, Gwo-Hsiung Tzeng. Combining neuralnetwork model with seasonal time series ARIMA model[J]. Technological Forecas ting & Social Change, 2002, 69: 71-87
[55] LeanYu, Shouyang Wang, K. K. Lai. A novel nonlinear ensemble foreca sting model incorporating GLAR and ANN for foreign exchange rates[J]. Computer s & Operations Research, 2005, 32: 2523-2541
[56]周开利,康耀红.神经网络模型及其MATLAB仿真程序设计[M].北京:清华大学出版社. 2005
[57]彭冬初.外汇投资一本通[M].北京:机械工业出版社,2008:105