This is an R package for Ron Begleiter's VMM Java code. 
http://www.cs.technion.ac.il/~ronbeg/vmm/index.html

It is basically undocumented, unless you already know about his code, or Variable 
order Markov Models in general. It is not fully tested, but it appears to work on 
toy problems.
It is an online learning algorithm, so you can add more sequences to the trained 
objects as you receive them.

It requires you have rJava package installed and working with Java (there may be 
path issues; don't blame me -blame Java). 

If you've never used an R package in raw form before, the way to make them is as 
follows: 

git clone https://github.com/locklin/VMM.git
R CMD build VMM
R CMD install VMM_version.tar.gz

From there, you should be able to load VMM with require(VMM) at your R command 
line.

ToDos:

I may encapsulate these simple functions as an S3 class to make CRAN people 
or OO nerds happy, but it's useful as it is for now for experiments with 
online learning.

Other To-Dos: gizmos to turn floating point numbers into symbol strings, 
altering the Java methods to allow some introspection into the learned 
objects.

It could use a simple test data set as well.