collaborative filtering - What are some ways for a reccomendation engine to deal with one time, novel and potentially important content? -
Say you have created a recommended engine that will show you a TV show to watch. For regular shows, using collaborative filtering and choice you can do a great job but say that it was something like Moonlanding in 1969. This is clearly an important event, you want your recommendation engine to handle that matter. But you can not rely on previous behavior, because when the show is over, the value of that recommendation becomes zero.
What are some effective ways to deal with this problem in the recommended place?
The problem in the CF is usually the opposite: Items without clicks / rating new < / Em> So far a CF algorithm can not be recommended and therefore users are having trouble. An old, well-known item should be easily suggested.
There is another problem: Some recommendation system algorithms support famous items, which may have better recommendations in everyone about more long tail, less-known objects, in fact some recommendations.
It seems as if you have the impression that this item is extra-good in some sense, the information of that side is that you can include the rated value added by a certain amount. I think the effective approach is just something like that.
Comments
Post a Comment