Application of Hidden Markov Model on the Prediction of Hepatitis B Incidences.

In this study, we apply a hidden Markov model (HMM) to the hepatitis B incidences series published by Chinese Center for Disease Control and Prevention. A two-univariate normal distribution is specified and estimated, where the number of states of the Markov chain is implied by maximum likelihood estimation. These two states, corresponding to different distribution laws, are interpreted as low incidence state and high incidence state accordingly. The probability of state transition is positive, albeit small. The historical states series can be inferred from the estimated HMM. We find that hepatitis B incidence is in low incidence state currently. Based on the estimation result, we predict that hepatitis B incidence will be in low incidence state in the future ten years.