Why we convert rating (1 to 5 or 1 to 10) to Binary Rating System for Collaborative Filtering Recommender Systems and what is benefit
2
-
$\begingroup$ Hi @OussamaAlahoum, welcome to the site. Can you provide more context about where you have seen this conversion applied? $\endgroup$– noeCommented Jun 14, 2023 at 10:29
-
$\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$– Community BotCommented Jun 14, 2023 at 16:04
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
You do not have to convert rating to the binary value to use in Collaborative Filtering. The result of the prediction will be the predicted rating for each user/item, and it can be in range 1 to 5, 1 to 10 or 0 to 1, depending on the data you provide. You, then, sort the output and recommend the items with the highest predicted ratings on top.
Here is an example from fastai collablearner
learn = collab_learner(
dls,
n_factors=50,
# here is the range of you ratings, you typically add 10% to the upper end
y_range=(1, 5.5)
)
learn.fit_one_cycle(
5,
0.01
)
