|
|
| Line 1: |
Line 1: |
| − | {{Projectproposal
| + | |
| − | |Contact person=Valentina Maccatrozzo
| |
| − | |Master areas=Multimedia, Information Sciences
| |
| − | |Project page=http://wiki.cs.vu.nl/mp/index.php/Tell_me_your_age_and_I%27ll_tell_you_what_you_like
| |
| − | |Fulfilled=No
| |
| − | }}
| |
| − | Standard collaborative filtering algorithms recommend programs to users taking into consideration the ratings of users with similar tastes. Given the ratings of the other users, these algorithms try to forecast the ratings of items that the target user didn't rate yet. However, sometimes it is simply not possible to identify the group of users with similar tastes of the target one, because enough information about her tastes are not available. This is usually called the cold start problem. We think that take into consideration some demographic information about the user can help in reducing this problem. For instance, take as target a group of women, between 30-35 years old, with at least one child. It is likely they watch similar TV programs. If a new user that fits this group starts to use our system, then we can already recommend her some programs based on the group preferences. This Master Project will investigate whether it is possible, at least partially, to fix the cold start problem. Three different approaches should be tested and combined in order to find the best combination of them: collaborative filtering (ratings), content based (items features) and demographics based.
| |
