Forecasting stock market based on price trend and variation pattern

  • Authors:

  • Ching-Hsue Cheng;Chung-Ho Su;Tai-Liang Chen;Hung-Hsing Chiang

  • Affiliations:

  • Department of Information Management, National Yunlin University of Science and Technology, Touliu, Yunlin, Taiwan;Department of Information Management, National Yunlin University of Science and Technology, Touliu, Yunlin, Taiwan and Department of Digital technology and Game Design, Shu-Te University, Yen Chau ...;Department of Information Management and Communication, Wenzao Ursuline College of Languages, Kaohsiung, Taiwan, ROC;Department of Information Management, National Yunlin University of Science and Technology, Touliu, Yunlin, Taiwan

  • Venue:
  • ACIIDS'10 Proceedings of the Second international conference on Intelligent information and database systems: Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.01

Visualization

Abstract

Since people cannot predict accurately what will happen in the next moment, forecasting what will occur mostly has been one challenging and concerned issues in many areas especially in stock market forecasting. Whenever reasonable predictions with less bias are produced by investors, great profit will be made. As the emergence of artificial intelligence (AI) algorithms has arisen in recent years, it has played an important role to help people forecast the future. In the stock market, many forecasting models were advanced by academy researchers to forecast stock price such as time series, technical analysis and fuzzy time-series models. However, there are some drawbacks in the past models: (1) strict statistical assumptions are required; (2) objective human judgments are involved in forecasting procedure; and (3) a proper threshold is not easy to be found. For the reasons above, a novel forecasting model based on variation and trend pattern is proposed in this paper. To verify the forecasting performance of the proposed model, the TAIEX (Taiwan Stock Exchange Capitalization Weighted Stock Index) of the year 2000, is used as experimental dataset and two past fuzzy time-series models are used as comparison models. The comparison results have shown that the proposed model outperforms the listed models in accuracy and stability.