Consistency of the maximum likelihood estimator in seasonal hidden Markov models
In this paper, we introduce a variant of hidden Markov models in which the transition probabilities between the states, as well as the emission distributions, are not constant in time but vary in a periodic manner. This class of models, which we will call seasonal hidden Markov models (SHMMs), is particularly useful in practice, as many applications involve a seasonal behaviour. However, up to now, there is no theoretical result regarding this kind of model. We show that under mild assumptions, SHMMs are identifiable: we can identify the transition matrices and the emission distributions from the joint distribution of the observations on a period, up to state labelling. We also give sufficient conditions for the strong consistency of the maximum likelihood estimator (MLE). These results are applied to simulated data, using the EM algorithm to compute the MLE. Finally, we show how SHMM can be used in real-world applications by applying our model to precipitation data, with mixtures of exponential distributions as emission distributions.
KeywordsHidden Markov models Climate modelling Identifiability Maximum likelihood
Mathematics Subject Classification62P12
We thank the anonymous reviewer for his/her comments. The author would like to thank Yohann De Castro, Élisabeth Gassiat, Sylvain Le Corff and Luc Lehéricy from Université Paris-Sud for fruitful discussions and valuable suggestions. This work is supported by EDF. We are grateful to Thi-Thu-Huong Hoang and Sylvie Parey from EDF R&D for providing this subject and for their useful advice.
- Broniatowski, M., Celeux, G., Diebolt, J.: Reconnaissance de mélanges de densités par un algorithme d’apprentissage probabiliste. Data Anal. Inform. 3, 359–373 (1983)Google Scholar
- Cappé, O., Moulines, E., Rydén, T.: Inference in hidden Markov models. In: Proceedings of EUSFLAT Conference, pp. 14–16 (2009)Google Scholar
- Dempster, A.P., Laird, N.M., Rubin, D.B.: Maximum likelihood from incomplete data via the EM algorithm. J. R. Stat. Soc. Ser. B (Methodol.) 39, 1–38 (1977)Google Scholar
- Katz, R.W.: Use of conditional stochastic models to generate climate change scenarios. Clim. Change 32(3), 237–255 (1996)Google Scholar
- Rabiner, L., Juang, B.: An introduction to hidden Markov models. IEEE ASSP Mag. 3(1), 4–16 (1986)Google Scholar
- Wilks, D.: Multisite generalization of a daily stochastic precipitation generation model. J. Hydrol. 210(1), 178–191 (1998)Google Scholar
- Wu, C.J.: On the convergence properties of the EM algorithm. Ann. Stat. 95–103 (1983)Google Scholar
- Yaoming, L., Qiang, Z., Deliang, C.: Stochastic modeling of daily precipitation in China. J. Geogr. Sci. 14(4), 417–426 (2004)Google Scholar
- Yoon, B.J.: Hidden Markov models and their applications in biological sequence analysis. Curr. Genomics 10(6), 402–415 (2009)Google Scholar