Recomendations
Whatever your system is it isn't very good. I rated a movie a 1 star and you recommend the sequel? Why would i want to watch the sequel to a movie I hated? You need a complete overall over your system clearly.
Anonymous
shared this idea
While we won’t be overhauling our recommendation algorithm, we do hope to review the tuning of their parameters. Until them, please see the comment for advice on how your personal recommendations might be improved.
-
Howdy,
We've been hearing reports lately about strange behavior from our recommender, and reviewing the tuning of our algorithms globally is one of the things on our todo list.
Ultimately recommendation algorithms can be rather inscrutable, so I don't know that I will be able to exactly explain the example you describe, where a 1 star rating led to being recommended the sequel. I can say that it depends on which recommendation algorithm you are currently using (MovieLens offers 4) -- two of the recommendation algorithms are not highly responsive to user input (this is intentional -- they are designed for users as they enter the recommender system and do not have many ratings to base recommendations on) Of the two high-quality recommenders, one of them (the wizard) only updates it's model of your preferences nightly, so a rating might not immediately be reflected in your recommendations.
Other than giving the algorithm a little time to see if it better understands you after it's next major model rebuild (usually nightly) I can also recommend using the "Star menu" (it's next to the account menu on the top-right, and displays your number of rated movies next to it) this menu will allow you to switch recommendation algorithm, if one of them is not modeling your preferences well, you may find that a different algorithm models your preferences better.
Hope that helps you optimize your experience while we work on the backend to make sure our algorithms are globally tuned as well as possible.