碩士 / 萬能科技大學 / 經營管理研究所 / 97 / Past study in the Fuzzy time series prediction model for the history of Fuzzy time series data have the advantage and to establish a simple relationship between the fuzzy logic way of doing group forecast accurately the value of traditional time series methods in the develop-ment of prediction models is already quite complete and mature. But only in the precise value of the historical time series prediction and there is the risk of over-fitting.
This study will be combined with Fuzzy time series with the precise value of the tradi-tional advantages of time series. The fuzzy interval of time after the sequence of construction of a Fuzzy Generalized Autoregressive Conditional Heteroskedasticity (Fuzzy-GARCH) pre-diction model Fuzzy-GARCH then forecast generalization process. The sample is using the new system the Baltic Freight Index (BDI) the actual closing price during the week of data. It is from January 2006 until 1st week first in May 2009 only 1st week. It is total of 174 as the empirical research data. Fuzzy time series compared with the Fuzzy-GARCH's predictive ca-pability and index information to the BDI in order to verify the information.
The results showed that the samples in the same period of the forecast weeks forecast information on the BDI. RMSE of Fuzzy-GARCH is 210, MAE of Fuzzy-GARCH is 159, MAPE of Fuzzy-GARCH is 10.03%, they are compared with the fuzzy time series, RMSE of fuzzy time series is 285, MAE of fuzzy time series is 237, MAPE of fuzzy time series is 15.17%, had better predictive power in the Baltic index data validation. Fuzzy-GARCH time series compared with the fuzzy also verify have better predictive power. Fuzzy-GARCH value compared to traditional time-series model accurately define the scope of the prediction interval in order to comply with the logic of human thinking and decision making predictive capability in comparison to the fuzzy time series is a better predictive capability.
Identifer | oai:union.ndltd.org:TW/097VNU05457027 |
Date | January 2009 |
Creators | Chang-Che Wu, 吳昌哲 |
Contributors | Kang-Lin Chiang, 蔣岡霖 |
Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
Language | zh-TW |
Detected Language | English |
Type | 學位論文 ; thesis |
Format | 43 |
Page generated in 0.0237 seconds