主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2007, Vol. 15 ›› Issue (4): 105-110.

• 论文 • 上一篇    下一篇

考虑影响因素的隐马尔可夫模型在经济预测中的应用

张冬青, 宁宣熙, 刘雪妮   

  1. 南京航空航天大学经济与管理学院, 江苏 南京 210016
  • 收稿日期:2006-09-20 修回日期:2007-07-15 出版日期:2007-08-31 发布日期:2007-08-31
  • 作者简介:张冬青(1972- ),女(汉族),江苏泗洪人,南京航空航天大学经济与管理学院,博士研究生,研究方向:统计信号处理、时间序列预测等
  • 基金资助:

    国家软科学研究计划资助项目(2006GXQ3B203)

Application of Hidden Markov Model Considering Influencing Factors in Economic Forecast

ZHANG Dong-qing, NING Xuan-xi, LIU Xue-ni   

  1. College of Economics & Management, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
  • Received:2006-09-20 Revised:2007-07-15 Online:2007-08-31 Published:2007-08-31

摘要: 定量预测方法分为因果预测法和时间序列预测法,因果预测法利用预测变量与其他变量之间的因果关系进行预测,时间序列预测法是根据预测变量历史数据的结构推断其未来值。由于因果预测法只利用某个变量与其他变量之间的因果关系,但缺少描述变量自身时间序列结构的功能;而时间序列预测法只能描述变量自身序列的结构,但没有考虑其他相关因素的影响,因此本文提出基于观测向量序列的隐马尔可夫模型(HMM)预测方法,该方法能同时考虑变量自身序列结构以及相关因素的影响。首先介绍HMM基本理论;其次,在模型训练、隐状态序列估计的基础上,提出基于观测向量序列HMM预测算法;最后分别进行仿真实验和实证研究,结果表明该方法的有效性。

关键词: 隐马尔可夫模型, EM算法, Viterbi算法, 影响因素, 预测

Abstract: Quantitative forecasting methods can be divided into time series models and causal models.Causal models forecast by considering the effects of outside factors,while time series models attempt to predict the future values using historical data itself.However,time series models take into account the structure of historical data rather than the effects of causal factors,and causal models consider the effect of causal factors rather than the structure of history data Therefore,a forecasting method based on hidden Markov mo del(HMM) with multivariable data,which includes both the time series structure and causal factors,is proposed in this paper.Firstly,we introduce the basic theory of HMM;then the corresponding algorithm is developed after discussing model training and parameters estimation.At last,a simulation experiment and an empirical research are launched,and experimental results indicate that the model proposed is effective.

Key words: hidden markov model, expectation maximization algorithm, viterbi algorithm, causal factors, forecast

中图分类号: