Relevance
Hello everyone!
I do not find very relevant an indication based on 5 or 6 votes (which lacks consistency, given that the predicted score will probably change a lot as more users rate that particular movie).
My idea is to create a new indicator that should be named RELEVANCE.
I think the formula should be:
Confidence Level * Predicted Score = Relevance
That way a movie with a very high predicted score based on very few ratings would not be very relevant for me (yet) - it would rate low on relevance. It would not be relevant a movie that had lots of ratings, but a low predicted score - it would also rate low on relevance.
The most relevant recommendations are the ones that indicate HIGH PREDICTED SCORES based on MANY RATINGS.
(non-English speaker, sorry for that, but I hope I could make myself understood)
Keep up the great work!