# Copyright 2018 The Cornac Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ============================================================================
from math import ceil
from collections import defaultdict
from .base_method import BaseMethod
from .ratio_split import RatioSplit
from ..utils import get_rng
from ..utils.common import safe_indexing
[docs]
class StratifiedSplit(BaseMethod):
"""Grouping data by user or item then splitting data into training, validation, and test sets.
Parameters
----------
data: array-like, required
Raw preference data in the triplet format [(user_id, item_id, rating_value, timestamp)].
group_by: str, optional, default: 'user'
Grouping by 'user' or 'item'.
chrono: bool, optional, default False
Data is ordered by reviewed time or not. If this option is True, data must be in 'UIRT' format.
test_size: float, optional, default: 0.2
The proportion of the test set, \
if > 1 then it is treated as the size of the test set.
val_size: float, optional, default: 0.0
The proportion of the validation set, \
if > 1 then it is treated as the size of the validation set.
rating_threshold: float, optional, default: 1.0
Threshold used to binarize rating values into positive or negative feedback for
model evaluation using ranking metrics (rating metrics are not affected).
seed: int, optional, default: None
Random seed for reproducibility.
exclude_unknowns: bool, optional, default: True
If `True`, unknown users and items will be ignored during model evaluation.
verbose: bool, optional, default: False
Output running log.
"""
def __init__(
self,
data,
group_by="user",
chrono=False,
fmt="UIRT",
test_size=0.2,
val_size=0.0,
rating_threshold=1.0,
seed=None,
exclude_unknowns=True,
verbose=False,
**kwargs
):
super().__init__(
data=data,
fmt=fmt,
rating_threshold=rating_threshold,
seed=seed,
exclude_unknowns=exclude_unknowns,
verbose=verbose,
**kwargs
)
if group_by not in ["user", "item"]:
raise ValueError(
"group_by option must be either 'user' or 'item' but {}".format(
group_by
)
)
if chrono and (fmt != "UIRT" or len(self.data[0]) != 4):
raise ValueError(
'Input data must be in "UIRT" format for sorting chronologically.'
)
self.chrono = chrono
self.group_by = group_by
self.val_size = val_size
self.test_size = test_size
self._split()
def _split(self):
data = (
sorted(self.data, key=lambda x: x[3]) if self.chrono else self.data
) # sort data chronologically
grouped_indices = defaultdict(list)
for idx, (uid, iid, *_) in enumerate(data):
if self.group_by == "user":
grouped_indices[uid].append(idx)
else:
grouped_indices[iid].append(idx)
train_idx = []
test_idx = []
val_idx = []
for rating_indices in grouped_indices.values():
n_ratings = len(rating_indices)
n_train, _, n_test = RatioSplit.validate_size(
self.val_size, self.test_size, n_ratings
)
if self.chrono:
# training portion is chronologically sorted
# validation and test portions are randomly selected
rating_indices = (
rating_indices[:n_train]
+ self.rng.permutation(rating_indices[n_train:]).tolist()
)
else:
rating_indices = self.rng.permutation(rating_indices).tolist()
train_idx += rating_indices[:n_train]
test_idx += rating_indices[-n_test:]
val_idx += rating_indices[n_train:-n_test]
train_data = safe_indexing(data, train_idx)
test_data = safe_indexing(data, test_idx)
val_data = safe_indexing(data, val_idx) if len(val_idx) > 0 else None
self.build(train_data=train_data, test_data=test_data, val_data=val_data)
