Using the neural network model to predict taiwan dollar exchange rate

• Main Authors: SIN-MING SHU, 徐希銘
• Other Authors: tzung-tse liou
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/72410857430096188434

碩士 === 東吳大學 === 企業管理學系 === 91 === [摘要] Apart from the original sample test, Messe and Rogoff (1983) once analyzed movement in exchange rate by using structural model and time series in order to compare the accuracy in the exchange rate forecast. In result of that, they have discovered errors in the formula and sample test in this structural currency amylases model and therefore proved structural currency amylases model only valid in theory but not with real market forecast. Thus, performance in random walk model would certainly performed better than most of traditional structural model. In order to find out more advanced exchange rate forecast models and therefore resolve errors created by traditional structural model. We have decide to use Neural Network combined with traditional exchange rate forecast model as well as technical amylases on this experiment. Our aim for this experiment was to search the factors that contribute to the changes in exchange rate and therefore apply these factors as learning particles into the neural model so that we can predict long term TWD exchange rate with the ongoing amylases in the real sample test. Feeding BPN with real time TWD information together with the factors discovered Neural model test in order to forecast the accuracy, we have discovered that test itself was actually better performed than random walk model as well as regression test in forecast TWD movement.