Abstract
本論主要探討模糊時間序列，從介紹Q. Song及B.S. Chissom提出的模糊時間序列開始，瞭解模糊時間序列的理論架構、運算方式、預測步驟及其應用的範圍，並與傳統時間序列作比較。再根據已有的方法，提出新的模糊化方法，並介紹新方法的原理及運算方式。再針對模糊時間序列模式的因數及變數加以討論，並提出明確的預測步驟，如此可解決一些觀測值為語意值(linguistic values)的時間序列，且為預測方法增加新的思考空間。The proposes of this paper are to investigate the propertiesand methods of fuzzy time series. It starts off withintroducing the fuzzy time series analysis proposed by Q. Songand B.S. Chissom. Then, the theory structure, calculation,forecasting procedure and application of fuzzy time series arecharacterized from a comparison between traditional time seriesand fuzzy time series. Based on the pros and cons of Song andChissom's method,we propose a new method on fuzzification inwhich the theory and implementation are developed. In addition,to facilitate forecasting applications, the factors andvariables of a fuzzy time series model are discussed in detail.With these results, time series of which the observations arelinguistic values can then be carried out properly.