In this paper, we propose an Interactive hidden Markov model (IHMM). In a traditional HMM, the observable states are affected directly by the hidden states, but not vice versa. In the proposed IHMM, the transitions of hidden states depend on the observable states. We also develop an efficient estimation method for the model parameters. Numerical examples on the sales demand data and economic data are given to demonstrate the applicability of the model.
- Categorical time series
- Hidden Markov model
- Prediction of demand
- Steady-state probability distribution