Frequency-Weighted Fuzzy Time-Series Based on Fibonacci Sequence for TAIEX Forecasting
This paper proposes a new fuzzy time-series model for promoting the stock price forecasting, which provides two refined approaches, a frequency-weighted method, and the concept of Fibonacci sequence in forecasting processes. In empirical analysis, two dif
- PDF / 192,887 Bytes
- 8 Pages / 430 x 660 pts Page_size
- 60 Downloads / 211 Views
Abstract. This paper proposes a new fuzzy time-series model for promoting the stock price forecasting, which provides two refined approaches, a frequencyweighted method, and the concept of Fibonacci sequence in forecasting processes. In empirical analysis, two different types of financial datasets, TAIEX (Taiwan Stock Exchange Capitalization Weighted Stock Index) stock index and HSI (Hong Kong Heng Seng Index) stock index are used as model verification. By comparing the forecasting results with those derived from Chen’s, Yu’s, and Hurang’s models, the authors conclude that the research goal has been reached. Keywords: Fuzzy Time-series, Stock Price Forecasting, Fibonacci Sequence, Fuzzy Linguistic Variable.
1 Introduction Time-series models have utilized the fuzzy theory to solve various domain forecasting problems, such as temperature forecasting, university enrollments forecasting [1-6] and financial forecasting [7-12]. Especially in the area of stock price forecasting, fuzzy time-series models are often employed [10-12]. As Dourra (2002) notes, it is common practice to “deploy fuzzy logic engineering tools in the finance arena, specifically in the technical analysis field, since technical analysis theory consists of indicators used by experts to evaluate stock price.”[7] In stock technical analysis fields, Elliott (1938) proposed the Elliott Wave Principle which has been playing an important role in stock analysis for more than six decades [13-15]. The theory is closely related to time-series because it applies the Fibonacci sequence to predict the timing of stock price fluctuation. Therefore, this paper applies the sequence in time-series models to forecast stock prices. Lastly, based on the drawbacks of previous models, we propose a new fuzzy timeseries model and recommend two refined processes in the forecasting processes. By employing a ten-period of TAIEX(Taiwan Stock Exchange Capitalization Weighted Stock Index) and HSI (Hong Kong Heng Seng Index) stock index as experiment datasets, the verifications show that our model outperforms conventional and advanced fuzzy time-series models. T. Washio et al. (Eds.): PAKDD 2007 Workshops, LNAI 4819, pp. 27–34, 2007. © Springer-Verlag Berlin Heidelberg 2007
28
H.J. Teoh, T.-L. Chen, and C.-H. Cheng
The remaining content of this paper is organized as follows: Section 2 introduces the related literature of fuzzy time-series models; section 3, propose the new model and algorithm; section 4 evaluates the performance of the proposed model; and section 5 concludes this paper.
2 Related Works Fuzzy theory was originally developed to deal with the problems involving human linguistic terms [16-18]. Time-series models had failed to consider the application of this theory until fuzzy time-series was defined by Song and Chissom (1993) who proposed the definitions of fuzzy time-series and methods to model fuzzy relationships among observations [1]. In following research, Song and Chissom (1994) continued to discuss the difference between time-invariant and time-variant mode
Data Loading...
