100,480,507 ratings from 480,189 users across 17,770 movie titles collected between 1999 and 2005. The dataset includes customer IDs, movie IDs, star ratings on a 1-5 scale, and the specific date of each rating.
Use Cases
- Train collaborative filtering models to predict the Rating value for specific CustomerID and MovieID pairs
- Analyze temporal rating patterns and user fatigue using the Date column
- Evaluate matrix factorization techniques on a sparse matrix of 480,189 users and 17,770 items
Strengths
- Contains 100,480,507 discrete rating events with associated timestamps
- Includes a separate movie metadata file with MovieID, YearOfRelease, and Title
- Features a 1-5 integer rating scale for 17,770 unique film titles
- Covers a 6-year longitudinal period of user behavior from 1999 to 2005