A Framework for Time Series Forecasts
Dongqing Zhang, Yubing Han, Xuanxi Ning, Xueni Liu · 2008
In order to cope with the 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 hidden Markov model (HMM), is proposed in this paper. On the other hand, most of literatures about neural networks suppose that the number of input is invariable. Obviously, this assumption is improper in some cases. Therefore, sequential Monte Carlo (SMC) method is used for on-line selection of the input order. Firstly, a framework for time series forecasts based on RBF-HMM model is proposed. Secondly, an on-line prediction algorithm based on RBF-HMM model using SMC method is developed. At last, the data of weekly steel price are analyzed and experimental results indicate that the RBF-HMM model is effective.