Something like movielens, a project they eventually bought.
You do a simple k-means calculation of your friends or closest matches (if you have clustering or svm, I guess he didn't have) and then get predictions for free. Something like the Netflix recommender system, just very primitive.
You do a simple k-means calculation of your friends or closest matches (if you have clustering or svm, I guess he didn't have) and then get predictions for free. Something like the Netflix recommender system, just very primitive.