Interesting approach. I've been working on something similar to explore the lineage of submissions. Although my code won't find any dependencies not listed in the contest data like yours does (a nice touch, I might add), it will allow the user to compare code from different submissions.
This is an awesome entry. There are some great data mining techniques and insights here. The MATLAB coding contest is interesting in that it is a rapid open source project where every branch is captured. This shows how things evolve and good code is propagated.
Thanks, this is the kind of entry we were hoping for in designing this new contest.