Closed bkj closed 6 years ago
This is meant to recommend movies the viewer is likely to watch next. Hence you want it to be a function of both rating and timing. Hence we use the recent viewing history of the crowd as a basis for predicting what's next.
Predicting ratings alone would have it recommend any movie the viewer is likely to enjoy.
Predicting what's popular alone would only recommend the top movies being viewed, which is a decent coldstart, but we can and have done much better than that.
Scott
On Tue, Jun 5, 2018 at 7:52 PM, Ben Johnson notifications@github.com wrote:
In the samples/movielens example, it looks like we're predicting the timestamp -- is that right? Seems like a more typical example would be predicting the ratings?
$ head ml-20m_ratings.csv userId,movieId,rating,timestamp 1,2,3.5,1112486027 1,29,3.5,1112484676 1,32,3.5,1112484819 1,47,3.5,1112484727 1,50,3.5,1112484580
$ head -n 1 ml-20m_rating 1 2,1112486027:29,1112484676:32,1112484819:47,1112484727:50
~ Ben
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/amzn/amazon-dsstne/issues/175, or mute the thread https://github.com/notifications/unsubscribe-auth/ARNK9v3Pl2zXmLQk_XfD79FAcdSipfLlks5t50PjgaJpZM4Ub5mW .
OK -- makes sense I think. Thanks!
In the
samples/movielens
example, it looks like we're predicting the timestamp -- is that right? Seems like a more typical example would be predicting the ratings?~ Ben