基于贝叶斯网络的混沌时间序列预测  

Chaotic time series prediction based on Bayesian network

作　　者：朱原媛[1] 杨有龙[1] 张恒伟[1] 

机构地区：[1]西安电子科技大学理学院,西安710071

出　　处：《计算机工程与应用》2012年第13期100-104,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61075055)

摘　　要：混沌时间序列预测是混沌理论的一个重要方向和研究热点,在气象、水力、经济和通信等各个领域有着广泛的应用。然而,由于混沌时间序列高度复杂的非线性特征,很难从理论上定量研究。利用贝叶斯网络(BNs)在处理不确定知识方面的优势,并结合相空间重构理论,建立了混沌时间序列非线性全局预测模型,实现对其动力学特性分析,从而达到预测目的。实验结果表明:模型具有良好的稳定性和预测能力,并能够有效地克服过拟合现象。Chaotic time series prediction is a very important research area and hotspot of chaotic theory; it is widely used in weather, electric power, economy, commutation etc. However, because of the high complexity of the nonlinear character of chaotic time series, it is hard to make quantitative research in theory. This paper uses the advantages of Bayesian Networks （BNs） in dealing the uncertainty together with the theory of phase-space reconfiguration to build a nonlinear prediction model for the prediction of chaotic time series. In this way, it can analyze the dynamic character and realize the prediction. The experimental results show that this prediction model has good predictability and stability, and overcomes the over fitting effectively.

关 键 词：混沌时间序列 贝叶斯网络 预测 相空间重构 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]