The hidden Markov model (continued)
With use of this HMM architecture a state sequence (i.e. a prediction) can be generated as follows: first a state is chosen according to the initial state probabilities. Every following state is chosen according to the transition probabilities of the present state. The aim is to maximize the product of these probabilities and the emission symbol probabilities along the given sequence. It can be solved by the Baum-Welch algorithm.