Adaptive-expectation based multi-attribute FTS model for forecasting TAIEX

  • Authors:

  • Jing-Wei Liu;Tai-Liang Chen;Ching-Hsue Cheng;Yao-Hsien Chen

  • Affiliations:

  • Department of Information Management, National Yunlin University of Science and Technology, 123, Section 3, University Road, Touliu, Yunlin 640, Taiwan, ROC;Department of Information Management and Communication, Wenzao Ursuline College of Languages, 900 Mintsu 1st Road, Kaohsiung 807, Taiwan, ROC;Department of Information Management, National Yunlin University of Science and Technology, 123, Section 3, University Road, Touliu, Yunlin 640, Taiwan, ROC;Department of Information Management, National Yunlin University of Science and Technology, 123, Section 3, University Road, Touliu, Yunlin 640, Taiwan, ROC and Department of Information Managemen ...

  • Venue:
  • Computers & Mathematics with Applications
  • Year:
  • 2010

Quantified Score

Hi-index 0.09

Visualization

Abstract

In recent years, there have been many time series methods proposed for forecasting enrollments, weather, the economy, population growth, and stock price, etc. However, traditional time series, such as ARIMA, expressed by mathematic equations are unable to be easily understood for stock investors. Besides, fuzzy time series can produce fuzzy rules based on linguistic value, which is more reasonable than mathematic equations for investors. Furthermore, from the literature reviews, two shortcomings are found in fuzzy time series methods: (1) they lack persuasiveness in determining the universe of discourse and the linguistic length of intervals, and (2) only one attribute (closing price) is usually considered in forecasting, not multiple attributes (such as closing price, open price, high price, and low price). Therefore, this paper proposes a multiple attribute fuzzy time series (FTS) method, which incorporates a clustering method and adaptive expectation model, to overcome the shortcomings above. In verification, using actual trading data of the Taiwan Stock Index (TAIEX) as experimental datasets, we evaluate the accuracy of the proposed method and compare the performance with the (Chen, 1996 [7], Yu, 2005 [6], and Cheng, Cheng, & Wang, 2008 [20]) methods. The proposed method is superior to the listing methods based on average error percentage (MAER).