You can use Google search results as a Markov chain (with the ranks translated to probabilities). Then your chain is fitted to all of the visible web. Another

You can use Google search results as a Markov chain (with the ranks translated to probabilities). Then your chain is fitted to all of the visible web. Another cool use for ML is for categorical and textual data. A fitted Markov Chain can give the probability of a string occuring. Strings that are more random (such as spammy text) get a low probability. Strings that are similar (such as certain user agents) get a similar probability, without having to directly compare these.