- 學位論文
渾沌理論於美元匯率預測上之應用 : 以多度空間向量預測之研究
Forecasting Foreign Exchange Rate with Chaos Theory :A Study of Multidimensional Vector Forecasting
摘要
對亞洲各國而言，如何正確的預測外匯走向是一個非常重要的課題，尤其對於以國際貿易為主要經濟活動的國家 ─ 中華民國台灣而言，其重要性更是不言可喻。不論是中華民國的政府、企業或投資人而言，大家對匯率變動的走勢都非常關心；但是在一個越來越自由的開放式金融市場中，匯率的走向已非由單一機構或國家所能掌控。加上影響匯率變動的因素多的不計其數，使得正確估計匯率的走勢似乎遙不可及。由於傳統的線性匯率定價理論經常與實際匯率的走向不符，使得近年來預測匯率的各種非線性理論盛行於世，如選擇權理論及各種類神經網路理路，最近又有學者探討匯率走勢是否符合渾沌理論，但很少學者經由渾沌理論來預測外匯走勢。本研究首先探討新台幣兌美元之匯率變動是否存在渾沌現象?在證實有混沌現象後；再進一步驗證以渾沌理論預測新台幣兌美元匯率之績效。 本研究以三種統計方法來驗證新台幣兌美元之匯率變動，實證結果發現新台幣兌美元之匯率確實存在渾沌現象，接著再以渾沌理論為基礎進行匯率預測，本研究以多度空間向量法來分析匯率變動資料，並以相空間鄰近相似法進行匯率預測，實證結果發現渾沌理論對短期匯率預測之績效相當良好，但對長期預測之績效並不顯著，此結果符合渾沌理論中之對初始條件敏感之特性，此特性即一般所謂的“蝴蝶效應”。 最後，本研究建議政府、投資人及進出口企業能善用渾沌理論來進行匯率預測，以降低於金融市場交易的風險。
並列摘要
Accurate predicting foreign exchange movement is an important topic for all Far Eastern countries, especially for an international trading base economy of Taiwan. During the era of the 21st century, the foreign exchange movement in the free economic system can not be fully governed or controlled by any single financial institute or government. Therefore, the trend and movement of foreign exchange are extremely difficult to forecast accurately. Owing to the fact that the forecasting performance on exchange prediction of traditional linear theory had not been performed well in this 21st century, accurate prediction of foreign exchange rate via non-linear models has been examined by governments, academics and practitioner. The famous is and most popular nonlinear models are option pricing model and various neural network models. The motivation of this study is to explore a new method of accurate prediction of the movement of foreign exchange rate accurately. Consequently, this investigation focuses on the movement of Taiwanese dollar vs. US dollar’s forecasting performance based on Chaos theory. This work uses three statistical methods to investigate whether the exchange rate’s movement fits the chaotic phenomenon. The findings indicate that there is indeed a chaotic phenomenon occurred within the foreign exchange rate data. Therefore, forecasting foreign exchange rate based on the Chaos theory has been applied. This study analyzes data using multidimensional vector analysis and predicts foreign exchange rate via Phase Space Neighbor Number Method. The empirical result shows that the Phase Space Neighbor Number Method performs well for short-term forecasting, yet it doesn’t work well for long-run prediction. Due to the sensitivity to the initial conditions of Chaos theory, the result demonstrates that this work confirms the “butterfly effect”. Finally, this study suggests that the government, the international enterprises and the investors to apply the Chaos theory to forecast the foreign exchange rate in order to reduce the transaction risk.