def test_amazon_office(self):
random.seed(time.time())
if random.random() > 0.8:
ratings = office.load_feedback()
contexts = office.load_graph()
self.assertEqual(len(ratings), 53282)
self.assertEqual(len(contexts), 108466)
"""Fit to and evaluate PCRL [1] on the Office Amazon dataset.
[1] Salah, Aghiles, and Hady W. Lauw. Probabilistic Collaborative Representation Learning\
for Personalized Item Recommendation. In UAI 2018.
"""
from cornac.data import GraphModality
from cornac.eval_methods import RatioSplit
from cornac.experiment import Experiment
from cornac import metrics
from cornac.models import PCRL
from cornac.datasets import amazon_office as office
# PCRL jointly models user preferences and performs deep item features learning from auxiliary data (e.g., item relations).
# The necessary data can be loaded as follows
ratings = office.load_feedback()
contexts = office.load_graph()
# Instantiate a GraphModality, it make it convenient to work with graph (network) auxiliary information
# For more details, please refer to the tutorial on how to work with auxiliary data
item_graph_modality = GraphModality(data=contexts)
# Define an evaluation method to split feedback into train and test sets
ratio_split = RatioSplit(data=ratings,
test_size=0.2,
rating_threshold=3.5,
exclude_unknowns=True,
verbose=True,
item_graph=item_graph_modality)
# Instantiate PCRL
pcrl = PCRL(k=100, z_dims=[300], max_iter=300, learning_rate=0.001)
# See the License for the specific language governing permissions and
# limitations under the License.
# ============================================================================
"""Fit to and evaluate MCF on the Office Amazon dataset"""
from cornac.data import GraphModality
from cornac.eval_methods import RatioSplit
from cornac.experiment import Experiment
from cornac import metrics
from cornac.models import MCF
from cornac.datasets import amazon_office as office
# MCF leverages relationships among items, it jointly factorizes the user-item and item-item matrices
# The necessary data can be loaded as follows
ratings = office.load_feedback()
item_net = office.load_graph()
# Instantiate a GraphModality, it make it convenient to work with graph (network) auxiliary information
# For more details, please refer to the tutorial on how to work with auxiliary data
item_graph_modality = GraphModality(data=item_net)
# Define an evaluation method to split feedback into train and test sets
ratio_split = RatioSplit(data=ratings,
test_size=0.2, rating_threshold=3.5,
exclude_unknowns=True, verbose=True,
item_graph=item_graph_modality)
# Instantiate MCF
mcf = MCF(k=10, max_iter=40, learning_rate=0.001, verbose=True)
# Evaluation metrics