Yeah that's basically right. It lists the user's top ratings first, recommendations second. Recommendations are based on all ratings in the model though, in theory.
Yes the example uses slope one internally. I like this algorithm best as a general default. It is easy to tweak the example to use something else. It is just another algorithm - works differently from the canonical recommender sketched in the documentation and no doubt gives different results. On Mar 18, 2009 5:42 PM, "Grant Ingersoll" <[email protected]> wrote: Just to clarify, when I run the group lens demo, say with: http://localhost:8080/mahout-taste-webapp/RecommenderServlet?userID=12&debug=true I get something like: User:User[id:12] Recommender: GroupLensRecommender[recommender:CachingRecommender[recommender:SlopeOneRecommender[weighted:true, stdDevWeighted:true, diffStorage:MemoryDiffStorage]]] Top 20 Preferences: 5.0 111 Taxi Driver (1976) Drama|Thriller 5.0 1198 Raiders of the Lost Ark (1981) Action|Adventure 5.0 1221 Godfather: Part II The (1974) Action|Crime|Drama 5.0 2804 Christmas Story A (1983) Comedy|Drama ... Recommendations: 7.0 557 Mamma Roma (1962) Drama 5.5 2510 Just the Ticket (1999) Comedy|Romance 5.5 3295 Raining Stones (1993) Drama ... Thus, what I am seeing here, is the preferences are the movies that user 12 rated and which were then used to derive the recommendations listed at the bottom, right? Also, it says I'm using a SlopeOneRecommender, what does that do compared to the other recommenders, as discussed in http://localhost:8888/taste.html#examples? That is, is it appropriate to substitute in the other ones for this demo? Would it be worth updating the demo to show the results of the various recommenders? Thanks Grant
