I wasn't able to investigate everything I had planned on, mostly because it took longer than I expected. I may try and work on it after I get out of school if I have time.
I did try a couple modifications to BRISMF. First was using a user based neighborhood correction instead of item based. This gave a little improvement, but pretty much trivial at the cost of about 4 hours additional prediction time. I tried a couple of other things, but to no better success.
But I did end up with a good idea of how these algorithms work, and how recommender systems in general behave. Hopefully I can work a little more on this later.
Posts
