Absolutely! Here's an Ars Technica article on the subject. I think there's room in particular for more pattern analysis, most likely through correlation, clustering, or latent class analysis. The most common programs I've seen (and my view is limited) are R (for statistical analysis and visualization) and Python (for network analysis).

This kind of work falls under the umbrella of digital humanities, which also describes such things as digitizing texts, creating 3D models, and entering metadata for all sorts of things. A search for 'digital humanities' turns up all kinds of results: a chronology (mostly outlining major programs and funding initiatives), a Harvard course on the digital humanities, and a post on five featured scholars using digital approaches in different fields. I like this last site, since it gives a good feel for what's possible and being done, but there's also a site that generates a definition of the digital humanities from a database of such definitions, and I think this reveals the state of the field from a very different perspective.

The primary difficulty for these kinds of studies is that people with the programming skills to dabble in machine learning aren't likely to dabble in history (at least not to a meaningful extent) while historians aren't likely to have the knowledge or the opportunity to pursue machine learning. Nonetheless, there's a fair amount of work out there if you know how to search for it. If you search for 'digital humanities' 'machine learning', you should get a fair number of relevant hits.