Skip to main content

General vs one-class CF

Short answer

General collaborative filtering uses explicit multi-level ratings with known negatives, focusing on rating error. One-class collaborative filtering uses only positive implicit feedback, treating missing data as ambiguous, and evaluates with ranking measures like precision at k.

Compare general collaborative filtering Vs One-class Collaborative filtering.

General collaborative filtering uses explicit, graded feedback such as ratings from 1 to 5. One-class collaborative filtering (OCCF) works where only positive implicit feedback (clicks, purchases, likes) is observed.

BasisGeneral CFOne-class CF
Feedback typeExplicit ratings (many classes)Implicit, only positive (1)
Negative examplesKnown (low ratings)Not available
Missing entriesNot rated; ignoredMixture of dislike and unknown; must be handled
Preference strengthKnown from rating valueEstimated from frequency or count of actions
Data exampleMovie ratings 1 to 5Product purchases, page views
Data volumeUsers must make effort to rate, so sparseEasy to collect, more plentiful but noisy
MethodsNeighbourhood, matrix factorization on observed ratingsWeighted matrix factorization, negative sampling, ranking methods
EvaluationRating error (RMSE, MAE)Ranking measures (precision@k, recall, AUC)
ExampleNetflix star ratingsAmazon purchase history

✓ Answer checked by Claude Opus 5.5 (high effort)

All Information System chapter 7 questions →