基于合作对策的股指时间序列智能组合预测研究  

An Intelligent Combination Prediction for Financial Time Series Based on the Cooperative Game

作　　者：罗洪奔[1,2] 游达明[1] 

机构地区：[1]中南大学商学院,湖南长沙410012 [2]湖南大学科学技术研究院,湖南长沙410082

出　　处：《湘潭大学学报（哲学社会科学版）》2014年第6期68-73,78,共7页Journal of Xiangtan University:Philosophy And Social Sciences

摘　　要：针对我国股市呈现出非线性、大幅度、频繁波动的特征,提出一种基于合作对策的股指时间序列智能组合预测方法,利用改进GRACH方法从时间关联性的角度辨识股指时间序列;同时利用神经网络方法从多种经济指标之间的相关性出发,建立股指时间序列预测模型,同时引入合作对策方法,对两种预测方法进行组合。实证分析结果表明,采用本算法能够有效地将预测精度控制在2.68%以内,同时在RMSE、MAPE和F三项指标上,较单一模型有极大提升。In view of the characteristics of nonlinear, large amplitude, and frequent fluctuations in our country＇s stock market, this paper presents a prediction method of intelligent combination of financial time series based on cooperative game. The improved GRACH method is used to establish the stock index time series model from the time association; at the same time, the neural network method based on the correlation is used to establish the stock index time series forecasting model with various economic indicators. The cooperative game method is introduced in this paper to combine the forecasting methods. The empirical results show that, using the proposed algorithm, the prediction accuracy can be controlled effectively within 2.68%, at the same time, the indicators of RMSE, MAPE and F have greatly improved than the single model.

关 键 词：股指时间序列 组合预测 合作对策 

分 类 号：F201[经济管理—国民经济] F832.5