Thursday, January 10, 2008

Back from the holidays. Found a great new way to calculate global effects - giving me a score of .9507, better than the original Netflix collaborative filtering method at the beginning of the contest. Now can I use it to improve my overall score?

1 comment:

George said...

Congratulations for 8.01% improvement.

I admire the way you have attacked the problem. You seem to have taken a really different path in comparison to the other contestants.

I'm really curious if you use some of the "well known" algorithms like KNN, SVD... or use blending.

If you care to comment about the technical part of you work (how many algorithms and stuff) I would appreciate it.

Thanks anyway.