基于RBF-HMM模型的时间序列在线预测

Time series on-line prediction based on RBF-HMM model

针对非线性非高斯时间序列,提出观测噪声服从隐马尔可夫模型(HMM)的径向基函数(RBF)神经网络预测模型—RBF-HMM模型,该模型具有如下两个特点:(1)用隐节点数可变的RBF神经网络对时间序列进行非线性建模;(2)用HMM对非高斯噪声进行建模.并采用序列蒙特卡罗(SMC)方法实现RBF-HMM模型参数的动态调整和时间序列的在线预测.最后采用南京禄口国际机场日旅客吞吐量数据进行实证研究,结果表明该模型的有效性.

In order to cope with nonlinear and non-Gaussian time series,a RBF-HMM model,which is based on radial basis function(RBF) neural network with the assumption of measurement noise being a hidden Markov model(HMM) process,is proposed.The model has two characteristics as follows: 1) RBF neural networks with variable structure approximate nonlinear time series;2) Non-Gaussian noise is modeled by HMM.Furthermore,sequential Monte Carlo(SMC) method is used to adjust the parameters of RBF-HMM model dynamically and t...