Clustering Methods for Collaborative Filtering.pdf (Size: 737.47 KB / Downloads: 0)
The grouping of individuals into groups on the basis of the items that they have purchased
allows specific recommendations of new items for purchase:
If you and I have enjoyed many of the same movies, then I'll probably
enjoy other movies you like. Recommending blogs based
on the similarity of interests (collaborative filtering alias) is attractive
for many areas: books, CDs, movies, etc., but are not always
work well. Because data are still scarce - a particular person has
given that a small fraction of all movies - much more accurate forecasts
may be made by the grouping of people clustered with similar
films and movies from clustering that tend to be loved by
the same people. Find optimal clusters is delicate because the movie
groups should be used to help determine the groups of persons and visa
versa. We present a formal statistical model of collaborative filtering,
and compare the different algorithms for the estimation of the model parameters
including changes in K-means and Gibbs Sampling. This
formal model is easily extended to manage the grouping of objects with
multiple attributes.