maciejkula / recommender_datasets

A common format and repository for various recommender datasets.

Recommender datasets

Parses and packages popular recommender datasets as simple-to-use CSV and HDF5 files. Have a look at releases for download links.

recommender_datasets's Issues

Item features

I would like to extend for starters the Movielens datasets to also include item features besides the interactions (user_id, item_id, rating, timestamp). These could then also be used in Spotlight I guess. I would add an hdf5 group called items (later one could also have users for users' features) and create for each feature a dataset in there whereby item_id is the index for the corresponding item.
Since the the interactions datasets stay at the root of the hdf5 file this should not break compatibility with Spotlight.
What do you think?

Error in generating Movielens dataset

Hi, I just tried generating the Movielens dataset from the master by running python datasets but got:

Processing movielens_100K
Traceback (most recent call last):
  File "datasets", line 110, in <module>
    download_movielens()
  File "datasets", line 31, in download_movielens
    data_fnc())
  File "/Users/fwilhelm/Sources/recommender_datasets/recommender_datasets/output.py", line 71, in write_hdf5_data
    if not isinstance(data[0], np.ndarray):
TypeError: 'generator' object is not subscriptable

Do I need to call it in another way?