DataInfo

class libreco.data.DataInfo(col_name_mapping=None, interaction_data=None, user_sparse_unique=None, user_dense_unique=None, item_sparse_unique=None, item_dense_unique=None, user_consumed=None, item_consumed=None, user_unique_vals=None, item_unique_vals=None, sparse_unique_vals=None, sparse_offset=None, sparse_oov=None, multi_sparse_unique_vals=None, multi_sparse_combine_info=None, seed=42)

Object for storing and updating information of indices and features.

Parameters:

• col_name_mapping (dict of {dict : int} or None, default: None) – Column name to index mapping, which has the format: {column_family_name: {column_name: index}}. If no such family, the default format would be: {column_family_name: {[]: []}}

• interaction_data (pandas.DataFrame or None, default: None) – Data contains user, item and label columns

• user_sparse_unique (numpy.ndarray or None, default: None) – Unique sparse features for all users in train data.

• user_dense_unique (numpy.ndarray or None, default: None) – Unique dense features for all users in train data.

• item_sparse_unique (numpy.ndarray or None, default: None) – Unique sparse features for all items in train data.

• item_dense_unique (numpy.ndarray or None, default: None) – Unique dense features for all items in train data.

• user_consumed (dict of {int : list} or None, default: None) – All consumed items by each user.

• item_consumed (dict of {int : list} or None, default: None) – All consumed users by each item.

• user_unique_vals (numpy.ndarray or None, default: None) – All the unique users in train data.

• item_unique_vals (numpy.ndarray or None, default: None) – All the unique items in train data.

• sparse_unique_vals (dict of {str : numpy.ndarray} or None, default: None) – All sparse features’ unique values.

• sparse_offset (numpy.ndarray or None, default: None) – Offset for each sparse feature in all sparse values. Often used in the embedding layer.

• sparse_oov (numpy.ndarray or None, default: None) – Out-of-vocabulary place for each sparse feature. Often used in cold-start.

• multi_sparse_unique_vals (dict of {str : numpy.ndarray} or None, default: None) – All multi-sparse features’ unique values.

• multi_sparse_combine_info (MultiSparseInfo or None, default: None) – Multi-sparse field information.

• seed (int, default: 42) – Random seed.

Variables:

• col_name_mapping (dict of {dict : int} or None) – See Parameters

• user_consumed (dict of {int, list}) – Every users’ consumed items in train data.

• item_consumed (dict of {int, list}) – Every items’ consumed users in train data.

See also

MultiSparseInfo

property global_mean

Mean value of all labels in rating task.

property min_max_rating

Min and max value of all labels in rating task.

property sparse_col

Sparse column name to index mapping.

property dense_col

Dense column name to index mapping.

property user_sparse_col

User sparse column name to index mapping.

property user_dense_col

User dense column name to index mapping.

property item_sparse_col

Item sparse column name to index mapping.

property item_dense_col

Item dense column name to index mapping.

property user_col

All the user column names, including sparse and dense.

property item_col

All the item column names, including sparse and dense.

property n_users

Number of users in train data.

property n_items

Number of items in train data.

property user2id

User original id to inner id mapping.

property item2id

Item original id to inner id mapping.

property id2user

User inner id to original id mapping.

property id2item

User inner id to original id mapping.

property data_size

Train data size.

__repr__()

Output train data information: "n_users, n_items, data density".

assign_user_features(user_data)

Assign user features to this data_info object from user_data.

Parameters:

user_data (pandas.DataFrame) – Data contains new user features.

assign_item_features(item_data)

Assign item features to this data_info object from item_data.

Parameters:

item_data (pandas.DataFrame) – Data contains new item features.

property popular_items

A number of popular items in train data which often used in cold-start.

save(path, model_name)

Save DataInfo Object.

Parameters:

• path (str) – File folder path to save DataInfo.

• model_name (str) – Name of the saved file.

classmethod load(path, model_name)

Load saved DataInfo.

Parameters:

• path (str) – File folder path to save DataInfo.

• model_name (str) – Name of the saved file.

class libreco.data.MultiSparseInfo(field_offset: Iterable[int], field_len: Iterable[int], feat_oov: ndarray, pad_val: Dict[str, Any])

dataclasses for storing multi-sparse features information.

A group of multi-sparse features are considered as a “field”, e.g., (“genre1”, “genre2”, “genre3”) form a “genre” field, and features belong to the same field share the same oov.

Variables:

• field_offset (list of int) – All multi-sparse fields’ offset in all expanded sparse features.

• field_len (list of int) – All multi-sparse fields’ sizes.

• feat_oov (numpy.ndarray) – All multi-sparse fields’ oov.

• pad_val (dict of {str : Any}) – Padding value in multi-sparse columns.