Can we use Markov Chain for forecasting a time series? I have a dataset which contains information about monthly visitors go to a particular website from Google News. Basically there are two columns - one is Month and another is Unique Visitors. There are 30 observations in the dataset. Can I apply Markov Chain to the data in any manner? Will it give any good result?
 A: I suppose you can use a Markov Chain for forecasting. However, you only have 30 data points, just two years and a half. This is very little data to use such a complicated method. I'd suggest you start with far simpler methods, like Exponential Smoothing. This textbook gives a good introduction to forecasting.
A: A Markov chain process and a time series process are two completely different kinds of stochastic processes, though all methods behind stochastic processes have similar features. You should distinguish different stochastic processes first by looking at the following table, which is taken from juan2013integrating. [2]

You can try to convert a Markov chain process to a time series process by normalising, which makes the state space relatively continuous. But I don't see the point of any conversion.
Usually, time series analysis focuses on dynamics regarding many lags, while the main idea behind Markov chains is to omit the history and focus on the current state. Also, time series analysis has been applied widely, while theoretical approaches for other three kinds are emphasised most of the time.

[2]: Morales, J.M., Conejo, A.J., Madsen, H., Pinson, P. and Zugno, M., 2013. Integrating renewables in electricity markets: operational problems (Vol. 205). Springer Science & Business Media.
