CMStatistics 2017: Start Registration
View Submission - CFE
A1849
Title: State space approach to online learning and forecasting in mixture autoregressive model Authors:  Mohammad Reza Yeganegi - International Institute for Applied Systems Analysis (IIASA) (Austria) [presenting]
Rahim Chinipardaz - Shahid Chamran University of Ahvaz (Iran)
Abstract: Online learning (parameter estimation) and forecasting of time series models are of great interest in both dynamic control and forecasting financial markets. The focus is on online parameter estimation in mixture autoregressive time series using state space approach. The state space representation of mixture autoregressive is given and stability and steady state of this representation is investigated using simulation study. The EM and online EM algorithms is organized for parameter estimation in state space. The performance of proposed method is also investigated using simulation study.