隨著全球國際間貿易的自由化,各國之間資金流動日益頻繁,再加上布列敦森林制度(Bretton Woods Agreement)崩潰後,各國放棄舊有的固定匯率制度,而改採浮動匯率制度。在浮動匯率的制度下,造成匯率的不確定性,也為各國間的投資及貿易活動投入了變數。因此若是能了解匯率變動的因素,並且能夠掌控匯率的變動,便可以降低匯率變動所造成的損失及風險,並且作為相關資訊給予政府機關、銀行業、國際企業及投資者在匯率上之參考。 本研究利用粗糙集理論(Rough Set Theory)結合窮舉演算法(Exhaustive Algorithm)與基因演算法(Genetic Algorithm)來建立決策規則並加入縮減率來探討預測準確率,並與局部轉換函數分類方法(Local Transfer Function Classifier;LTF-C)進行比較,分別對美元、日圓、港幣、加幣、韓幣、泰銖兌台幣為研究對象,最後找出影響匯率變動的因素。實證結果發現,粗糙集結合窮舉演算法與基因演算法皆優於LTF-C,在韓幣、日圓、泰銖的預測準確率上以窮舉演算法與基因演算法有較佳的表現,此外基因演算法當規則縮減率為0.9時,各國預測準確率皆為最高;而窮舉演算法並無此現象。
Because of the global liberalization of international trade, fund between nations flows more frequently day by day. Besides, after the collapse of Bretton Woods Agreement, many countries had given up the fixed exchange rate system and used the floating exchange rate system instead. Under the floating exchange rate system, the prediction of actual exchange rate becomes uncertainty. Therefore, if one could fully understand and control the factors that caused the exchange rate movements, we would able to reduce the losses caused by exchange rate fluctuations and risks, and provide some relevant information for government agencies, banks, international companies and investors in the exchange rate of reference. In this study, we are engaged in predicting monthly exchange rate of New Taiwan dollar to U.S. dollar, Japanese Yen, Hong Kong dollars, Canadian dollars, Korean won and Thai baht based on Rough Set Theory combined with Exhaustive Algorithm and Genetic Algorithm, building the decision rules and adding shortening ratio to investigate the prediction accuracy, and finally we also compare the prediction performance with Local Transfer Function Classifier. According to the empirical results, the prediction performance of Exhaustive Algorithm and Genetic Algorithm outperform Local Transfer Function Classifier. In the Korean won, Japanese yen and Thai baht, the forecasting accuracy under Exhaustive Algorithm and Genetic Algorithm has better performance. In addition, the performance of Genetic Algorithms under the shortening ratio of 0.9 has the highest prediction accuracy among all countries; while this phenomenon doesn’t exist if we use by Exhaustive Algorithm.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。