EM-algorithm for training of state-space models with application to time series prediction

In this paper, an improvement to the E step of the EM algorithm for nonlinear state-space models is presented. We also propose strategies for model structure selection when the EM-algorithm and statespace models are used for time series prediction. Experiments on the Poland electricity load time series show that the method gives good shortterm predictions and can also be used for long-term prediction.

• Type: Conference
• Year: 2006
• Where: ESANN
• Authors: Elia Liitiäinen, Nima Reyhani, Amaury Lendasse
• Keywords: ESANN 2006, ESANN 2007, Load Time Series, Nonlinear State-space Models, Time Series Prediction