An application of Poisson hidden Markov model to forecast endometrial cancer cases in Odisha, India

Background: Time series model often used for forecasting, but inadequate when the data series is unbounded and over dispersed in nature. Moreover, if the observations are serially dependent, then Markov dependent mixture model i.e., Poisson Hidden Markov model can be used. The objective of this study was to apply Poisson Hidden Markov model to forecast month wise new endometrial cancer cases. Materials and Methods: Month-wise total number of registered endometrial cancer cases has been collected from a local cancer hospital in Odisha from January, 2017 to December, 2021. In this paper we have applied the Poisson hidden Markov model to forecast, the number of endometrial cancer cases for next 12 months. Results: Three state Poisson hidden Markov model was found as best fitted model and used to forecast the endometrial cancer cases for the next 12 months. The results showed that the number of endometrial cancer cases most likely to lie in state 2 in January and in state 3 in rest of the months in 2022. The monthly forecasted mean of endometrial cancer cases varies between 34 to 38 for the year 2022. Conclusion: This study reveals that the average endometrial cancer cases will increase in future months. It is also suggested that, the three-state hidden Markov model can be used to fit and forecast the distribution of the number of endometrial cancer cases.