I've just finished implementing a collaborative recommender system based on views, sales and similar user-product interactions. This approach works quite well and I am pleased with the results.
Nevertheless, I am want to know what influence the integration of some item specific features would have. Approaches like computing the similarity of user and item vectors within the same feature space (like term vectors of product descriptions) are easy to scale with MapReduce. I wonder how model based approaches might be scaled to a large number of users. My understanding is that I would have to train some model like a decision tree or naive bayes (or regression … etc.) for each user and do the prediction for each item using this model. Is there any common approach to get those techniques scaling up with larger datasets? Many thank in advance, Dominik