Recommendation Datasets

From RecSysWiki
Jump to: navigation, search

Datasets used for recommender systems research purposes come from various domains, in different sizes and with different means of expressing user actions. The table below shows a comparison of some of the more common datasets for recommendation.

Dataset Domain No. of user actions User actions No of users License
MovieLens 100k Movies 100 k ratings 1k custom
MovieLens 1M Movies 1 M ratings 6k custom
MovieLens 10M Movies 10 M/100k ratings/tag assignments 72k custom
MovieLens 20M Movies 20 M/645k ratings tag/assignments 138k custom
MovieTweetings Movies 200k ratings 25k custom
RARD: Related-Article Recommendations Research Articles 57m / 77k (non-)clicks 7m sessions (no users) CC-BY 3.0

See also