PHMM Based on Nonparametric Bayesian Estimation
Ouyang Chu-fe · Journal of Chinese Computer Systems · 2015
Traditional time series model barely analysis the time series data from a system angle. So far,a Pattern based Hidden Markov Model has implemented the inner state transition mechanism of the time series production system,which provided an effective way to address multi-step prediction and relative detection in time series. In this article,we propose a Nonparametric Bayesian Estimation approach to calculate the inner distribution of the hidden state. On the contrary of the fixed state Hidden Markov Model,PHMMis an infinite state Hidden Markov Model. By introducing the Hierarchical Dirichlet Process,we can explain the two level stochastic process of PHMM. Experimental results on authentic data indicate the effectiveness of the Nonparametric Bayesian Estimation combined with PHMM.