隸屬度函數及區間長度改良對模糊時間序列預測之影響探討

THE INFLUENCE OF FUZZY MEMBERSHIP FUNCTION AND THE IMPROVEMENT OF LENGTH OF INTERVAL ON THE FUZZY TIME SERIES FORECASTING

林廷臻(Ting-Jen Lin)

指導教授 ： 張建瑋

大同大學/電機資訊學院/應用數學學系所/碩士(2008年)

摘要

自1991年Song提出模糊時間序列模式來對具有不確定性之時間序列進行預測，後續有許多相關的研究，且都獲致不錯的結果。
然而，在建構模糊時間序列模型時，如何有效地偵測模型的階次、定義適當的隸屬度函數的區間長度、計算模糊自相關矩陣等，一直是相關研究爭議的重點，本文將針對隸屬度函數的區間長度的比較來進行探討並提出改進之方法。
本文研究重點包括：模糊隸屬度函數的定義與範圍區間長度上的取捨，模糊自相關矩陣計算的化簡及模糊時間序列模式之建構與預測等。我們首先給定模糊時間序列模式的概念與一些重要性質。再來，藉著詳細的演算比較這些區間大小不同的模糊時間序列模型，我們確立模糊時間序列分析方法的有效性。最後，使用所建構的模糊時間序列模式對未來進行預測，以驗證所建構模糊時間序列模式的效率性與實用性。

關鍵字

類神經網路 ； 模糊時間序列 ； 模糊隸屬度函數 ； 模糊關係

並列摘要

Fuzzy time series model was proposed by Song Q. in 1991, and it can forecast data effectively. Many researches indicate this method is good for forecasting time-invariant data.
Since the establishment of fuzzy time series model, several controversial issues such as how to determine the order of the model, how to define the length of interval of the fuzzy membership function and how to calculate the fuzzy auto-correlation matrix. We will discuss the improvement of the interval length of the fuzzy membership function in this thesis.
This thesis is to decide the interval length of fuzzy membership function, to reduce the operation of fuzzy auto-correlation matrix, to establish the fuzzy time series model and to forecast the fuzzy time series. First, we define some important definitions of fuzzy time series. Second, we compare the different fuzzy time series model and thus claim the effectiveness for the analysis of fuzzy time series. Finally, we forecast the data with the fuzzy time series model and prove the efficiency and practicability of these models.

並列關鍵字

fuzzy time series ； neural network ； fuzzy relation ； fuzzy membership function

參考文獻

[21] Ruey-Chyn Tsaur, Jla-Chi O Yang and Hslao-Fan Wang, “Fuzzy Relation Analysis in Fuzzy Time Series Model,” Computers and Mathematics with Applications, 49 (2004) 539-548.

[6] Berlin Wu and Min-Chih Liaw, “Application of Fuzzy Time Series Analysis to Change Periods Detection,” IEEE International Fuzzy Systems Conference Proceedings, (1999) 22-25.

[7] Berlin Wu and Shu-Lin Hung, “A fuzzy identification procedure for nonlinear time series: With example on ARCH and bilinear models,” Fuzzy Sets and Systems, 108 (1997) 275-287.

[8] Chao-Hung Wang, “Predicting tourism demand using fuzzy time series and hybrid grey theory,” Tourism Management 25 (2004) 367-374.

[9] Dug-Hun Hong, “A note on a fuzzy time series model of fuzzy number observations,” Fuzziness and Knowledge-Based Systems, 12 (2003) 1-5.