Yeah, the return value of estimatePreference() is meaningless in this context, since you have no preference values (or they're all, conceptually, the same). The result of other methods would still have meaning though, like recommend(). On Tue, Jul 28, 2009 at 3:59 PM, James James wrote: > I also built a recommender using taste which operates on a binary data set (an user has bought or not bought a product). However, the recommender always return the same predicted value for all test items (in other words, all test items are considered preferred by the recommender). Did I do something wrong? > > James > > >