Data sets CSV format

Data sets file are assumed of being CSV where each row is a user-item rating. Each row must be formed of at least two columns where the first column represents the user raw id and the second column the raw item id.

For example:

user_id,item_id
u1,i1
u1,i2
u2,i1
u2,i3

The example shows a data set CSV with 4 ratings, 2 users (u1, u2) and 3 items (i1,i2,i3). The CSV uses colon as delimiter and it has a header which must be configured correctly in the data configuration file (see Configuration files format). Since the data set has implicit feedback, in the configuration file it must be set topn to 1.

In the case of data sets for rating predition, a third column must be provided containing the rating value. For example:

user_id,item_id,rating
u1,i1,3.5
u1,i2,2
u2,i1,4
u2,i3,5

which is a data set with the same users as the one above, but with explicit ratings.