Heuristic Methods in Time Series Analysis Processing Using Fuzzy Statistics

應用模糊統計與時間數列轉折點之研究

10.30053/CHJM.200403.0004

斐文(Wei Pei)；鄭舜仁(Shuenn-Ren Cheng)

轉折點 ； 累加和法 ； 模糊熵 ； 遺傳偵測法 ； 決策法則 ； Fuzzy Statistics ； Fuzzy Entropy ； Change Period ； Detection and Identification ； Decision Making

中華管理學報

5卷1期（2004 / 03 / 01）

57 - 76

英文

本文主要探討時間序列轉折點的偵測程序與決策建構過程，我們應用模糊統計程序包括：累加移動平均和、CUSUM、模糊熵及遺傳偵測法來偵測轉折點，並比較其效力，這些方法較傳統的方法有些不同，主要是我們結合了統計檢定程序與模糊理論的最新觀念，在偵測程序過程中，我們藉應用模糊邏輯推論，提出了幾個實用演算法則，並用台灣的商業景氣循環指標為實證例子，以驗證我們所提出演算則的實效性，並作為今後決策者在於經濟模型、計量管制、財務決策方面提供更佳決策模式。

In this research, we are to investigate how heuristic methods using fuzzy statistical theory can be used in data processing and its application. Instead of merely trying to detect a turning point, in this paper, we construct several fuzzy procedures based on fuzzy statistics and AI concept for different case. The application of various detecting procedures can be found very useful in the many fields, such as, economic modeling, quality control, finance, biology, meteorology and ecology are discussed. We believe that these efficient approaches have advantages over model based criteria or other 'black-box' machine learning procedures in that it produces transparent decision models that are easily understood by decision-makers.