Detecting influenza epidemics using hidden Markov models with Bayesian approach

In this thesis, we present a statistical method for detecting influenza epidemics. First, we use a hidden Markov model with Bayesian approach to partition the influenza data into two groups, one group for the epidemic states and another one for the non-epidemic states. Then, we detect the start of the epidemic phase of the disease through introducing a warning threshold. This warning threshold is efficient in increasing the detection rates while decreasing the false alarm rates. Finally, we compare the established hidden Markov model with the traditional seasonal ARIMA model