基於變動猶豫程度之直覺模糊時間序列模型

在1986年時，Atanassov學者針對模糊集理論進行了擴展，提出直覺模糊集理論，對於處理不確定性的問題有一定的成效。在2007年時，Castillo學者等人首次將直覺模糊集合理論結合到時間序列的分析中，提升了預測精準度。因此，近幾年與直覺模糊時間序列的相關研究變得活絡。然而在與直覺模糊時間序列的相關研究中，大部份的研究皆以主觀意見決定猶豫程度，甚至部份研究在區間長度及語意個數的仍是以主觀意見去決定的，因此無法以客觀的角度反應出資料集的特性。有鑑於此，本研究的方向主要是探討在直覺模糊時間序列的分析中，如何以直覺模糊時間序列的最小歸屬值決定每個區間的猶豫程度、以資料筆數決定最適當的語意個數以及區間長度。本研究提出基於變動猶豫程度之直覺模糊時間序列模型，希望能透過該模型解決先前方法的缺失。為了驗證本研究的方法，將採用Alabama大學的入學人數、台灣經濟部能源局(Bureau of Energy, Ministry of Economic Affairs)所提供的台灣季總用電量以及台灣證券交易所(Taiwan Stock Exchange Corporation)所提供的台灣加權股價指數(Taiwan Stock Exchange Capitalization Weighted Stock Index, TAIEX)三組數據，並將本研究的模型以不同的績效指標以及條件與近年其他研究模式進行績效比較。而在研究結果顯示，本研究不僅改善先前研究的缺失，在預測績效也有進一步的提升。

關鍵字

直覺模糊時間序列；變動猶豫程度； N分位離散化方法；台灣加權股價指數

並列摘要

In 1986, Atanassov extended the fuzzy set theory and proposed the intuitionistic fuzzy set theory. It has an incremental effect for dealing with uncertainty problems. In 2007, the Castillo et. al combined the intuitionistic fuzzy set theory into the analysis of time series for the first time, which improved the prediction accuracy. Therefore, the researches related to intuitionistic fuzzy time series have become popular in recent years. However, in the studies related to intuitionistic fuzzy time series, most of them use subjective opinions to decide the degree of hesitation. Even some studies on the length of the interval and the number of linguistic are still decided by subjective opinions. Therefore, the characteristics of the data set cannot be reflected from an objective perspective. In view of this, the direction of this research is mainly to explore how to decide the degree of hesitation for each interval with the minimum attribution of intuitionistic fuzzy time series in the analysis of intuitionistic fuzzy time series. How to decide the most appropriate number of linguistic meanings and interval length by the number of data. This research proposes an intuitionistic fuzzy time series model based on the degree of variation hesitation and hopes that this model can solve the lack of previous methods through this model. To verify the methodology of this study, the number of enrolled at Alabama University, the total quarterly electricity consumption in Taiwan provided by the Bureau of Energy, Ministry of Economic Affairs, and the Taiwan Stock Exchange Corporation The Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) is provided with three sets of data, and the model of this study is compared with other research models in recent years with different performance indicators and conditions. The research results show that this study not only improves the lack of previous research but also further improves the predictive performance.