系统仿真学报 ›› 2018, Vol. 30 ›› Issue (6): 2086-2094.doi: 10.16182/j.issn1004731x.joss.201806011

• 仿真建模理论与方法 • 上一篇    下一篇

基于合作对策的股指时间序列组合预测

罗伟   

  1. 湖南铁道职业技术学院,湖南 株洲 412001
  • 收稿日期:2016-08-06 修回日期:2016-11-01 出版日期:2018-06-08 发布日期:2018-06-14
  • 作者简介:罗伟(1979-),男,湖南株洲,硕士,副教授,研究方向为控制工程。
  • 基金资助:
    湖南省教育厅科学研究项目(13C591)

Combination Forecasting of Stock Index Time Series based on Cooperative Game Theory

Luo Wei   

  1. Hunan Railway Professional Technology College, Zhuzhou 412001, China
  • Received:2016-08-06 Revised:2016-11-01 Online:2018-06-08 Published:2018-06-14

摘要: 针对我国股市呈现出非线性,大幅度、频繁波动的特征,提出一种股指时间序列智能组合预测方法。从多种经济指标之间的相关性出发利用神经网络方法,建立股指时间序列预测模型;同时从时间关联性的角度出发,利用改进ARIMA方法,建立辨识股指时间序列发展趋势和规律,通过引入合作对策方法,对两种预测方法进行组合。仿真结果表明,采用此算法能够有效的将预测精度控制在6.6%以内,与单一模型在RMSE、MAPE和F三项指标比较具有较大优势。

关键词: 股指时间序列, 组合预测, ARIMA, 神经网络, 合作对策

Abstract: In view of the characteristics of nonlinear, large amplitude, frequent fluctuations in China's stock market, a prediction method of intelligent composite stock index time series based on the cooperative game is presented. The prediction model of stock index time series is established by using neural network method based on the correlations among the various economic indicators, and the development trend and laws of stock index time series are established by using the improved ARIMA method. The two methods are combined by importing cooperative game method. Simulation results show that the prediction accuracy of the presented method is controlled within 6.6% effectively, which has greater advantage than the single model in the index evaluation such as RMSE, MAPE and F.

Key words: Financial time series, combination prediction, ARIMA, NN, cooperative game

中图分类号: