Forecasting Model for Chaotic Time Series with Seasonality

• Main Authors: Fu-Yuan Chen, 陳富源
• Other Authors: Ping-Teng Chang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/40906114323952195866

碩士 === 東海大學 === 工業工程與經營資訊學系 === 101 === For a seasonal chaotic time series, first we use the time-delay embedding theorem to reconstitute the sequence of phase space, which describes the behavior of a nonlinear system evolution, then we use the common chaotic prediction method: Full-area forecast method, Local-region forecast method, the weighted largest Lyapunov exponent forecasting method to do the chaotic time series forecasting. In order to improve the accuracy of chaotic prediction, we think the time series affected by seasonal factors, so we use the seasonal index adjustment methods, to further reduce the chaotic prediction errors. Some of chaotic time series will not show chaotic phenomena after adjusting seasonal index. As a result, we decided to use the seasonal index adjustment after phase space reconstruction. The results indicate both make the prediction of chaotic phenomena more precise and improve the prediction more as the changes are more obvious in the chaotic time series.