def test_predict(self):
alg = NXPageRank()
ratings = pd.DataFrame.from_records(
[("A000", "tt0114576", 0.5, "54654675"),
("A000", "tt0112453", -0.5, "54654675"),
("A001", "tt0114576", 0.8, "54654675"),
("A001", "tt0112896", -0.4, "54654675"),
("A000", "tt0113041", 0.6, "54654675"),
("A002", "tt0112453", -0.2, "54654675"),
("A002", "tt0113497", 0.5, "54654675"),
("A003", "tt0112453", -0.8, "54654675")],
columns=["from_id", "to_id", "score", "timestamp"])
try:
path = "../../../contents/movielens_test1591885241.5520566"
file = os.path.join(path, "tt0114576.xz")
with lzma.open(file, "r") as content_file:
pass
except FileNotFoundError:
path = "contents/movielens_test1591885241.5520566"
rank = alg.predict('A000', ratings, 1, path, ['tt0114576'])
logger.info('pg_rk results')
for r in rank.keys():
logger.info('%s %s', str(r), str(rank[r]))
def test_clean_rank(self):
rank = {
"A000": 0.5,
"tt0114576": 0.5,
"A001": 0.5,
"tt0113497": 0.5,
"tt0112453": 0.5
}
alg = NXPageRank(graph=graph)
# remove from rank all from nodes
result = alg.clean_rank(rank,
user_id="A000",
remove_profile=False,
remove_from_nodes=True)
expected = {"tt0114576": 0.5, "tt0113497": 0.5, "tt0112453": 0.5}
self.assertEqual(expected, result)
# remove from rank all from nodes and all data from user A000
result = alg.clean_rank(rank,
user_id="A000",
remove_profile=True,
remove_from_nodes=True)
expected = {"tt0113497": 0.5}
self.assertEqual(expected, result)
def test_extract_profile(self):
alg = NXPageRank(graph=graph)
result = alg.extract_profile("A000")
expected = {'tt0114576': 0.75, 'tt0112453': 0.25, 'tt0113041': 0.8}
self.assertEqual(expected, result)
def test_extract_profile(self):
alg = NXPageRank(graph=graph)
result = alg.extract_profile("3", graph)
expected = {"tt0112453": 0.55, "55117": 0.5, "M": 0.5}
self.assertEqual(expected, result)
def test_predict(self):
user_ratings = ratings[ratings['from_id'] == 'A001']
alg = NXPageRank()
rank = alg.predict(user_ratings, 2)
self.assertEqual(rank, {})
alg = NXPageRank(graph=graph)
rank = alg.predict(user_ratings, 1)
logger.info('pg_rk results')
for r in rank.keys():
print(str(r) + " " + str(rank[r]))
self.assertIn('tt0112453', rank.keys())
# alg = NXPageRank(graph=graph)
# rank_fs = alg.predict('A001', ratings, 1, feature_selection_algorithm=NXFSPageRank())
# logger.info('pg_rk results')
# for r in rank_fs.keys():
# print(str(r) + " " + str(rank_fs[r]))
alg = NXPageRank(graph=graph, personalized=True)
rank_personalized = alg.predict(user_ratings, 1)
logger.info('pg_rk results')
for r in rank_personalized.keys():
print(str(r) + " " + str(rank_personalized[r]))
self.assertIn('tt0113041', rank_personalized)
def test_clean_rank(self):
rank = {
"1": 0.5,
"tt0112281": 0.5,
"2": 0.5,
"tt0113497": 0.5,
"tt0112302": 0.5
}
alg = NXPageRank(graph=graph)
# doesn't remove any node
alg.remove_items_in_profile = False
alg.remove_properties = False
alg.remove_user_nodes = False
result = alg.clean_rank(rank, graph, "1")
self.assertGreaterEqual(len(result.keys()), 0)
# removes user and property nodes and item nodes already in the user profile
alg.remove_items_in_profile = True
alg.remove_properties = True
alg.remove_user_nodes = True
result = alg.clean_rank(rank, graph, "1")
expected = {"tt0113497": 0.5}
self.assertEqual(expected, result)
processor=NumberNormalizer(min_=1, max_=5))
],
from_field_name='user_id',
to_field_name='item_id',
timestamp_field_name='timestamp').import_ratings()
full_graph = NXFullGraph(
source_frame=ratings_import,
user_contents_dir=None,
item_contents_dir=item_contents_dir,
user_exogenous_properties=None,
item_exogenous_properties=['film_director', 'starring', 'producer'])
rank = NXPageRank(graph=full_graph).predict(
user_id='1',
ratings=ratings_import,
recs_number=10,
)
rank_pd = pd.DataFrame({
'from_id': ['1' for x in rank.keys()],
'to_id': [x for x in rank.keys()],
'rating': [x for x in rank.values()]
})
truth_rank = ratings_import[ratings_import['from_id'] == '1']
truth_rank = truth_rank.rename(columns={'score': 'rating'}, inplace=False)
print(truth_rank)
print(rank_pd)
def test_predict(self):
alg = NXPageRank(graph)
user_ratings = ratings[ratings['from_id'] == '2']
# test for number of recommendations <= 0
rank = alg.predict(user_ratings, -10)
self.assertEqual(len(rank), 0)
# test for standard prediction considering ratings from a user (PageRank with priors)
rank = alg.predict(user_ratings, 3)
self.assertEqual(len(rank), 3)
# test for prediction with empty dataframe (standard PageRank)
empty_ratings = pd.DataFrame()
rank = alg.predict(empty_ratings, 3)
self.assertEqual(len(rank), 3)
# test for prediction with a candidate_item_id_list
rank = alg.predict(user_ratings,
3,
candidate_item_id_list=['tt0113277', 'tt0114709'])
self.assertEqual(len(rank), 2)
# test for prediction with feature selection algorithms both for items and users and empty ratings
alg = NXPageRank(graph=graph,
item_feature_selection_algorithm=FSPageRank(1),
user_feature_selection_algorithm=FSPageRank(2))
rank = alg.predict(empty_ratings, 3)
self.assertEqual(len(rank), 3)
# test for prediction with feature selection algorithms both for items and users and user ratings
alg = NXPageRank(graph=graph,
item_feature_selection_algorithm=FSPageRank(1),
user_feature_selection_algorithm=FSPageRank(2))
rank = alg.predict(user_ratings, 3)
self.assertEqual(len(rank), 3)
# test for prediction with feature selection algorithm for items only
alg = NXPageRank(graph, item_feature_selection_algorithm=FSPageRank(1))
rank = alg.predict(user_ratings, 3)
self.assertEqual(len(rank), 3)
print(rank)
# test for prediction with feature selection algorithm for users only
alg = NXPageRank(graph, user_feature_selection_algorithm=FSPageRank(1))
rank = alg.predict(user_ratings, 3)
self.assertEqual(len(rank), 3)
print(rank)
# test for prediction with feature selection algorithm for items and users with k set to 0
alg = NXPageRank(graph,
item_feature_selection_algorithm=FSPageRank(0),
user_feature_selection_algorithm=FSPageRank(0))
rank = alg.predict(user_ratings, 3)
self.assertEqual(len(rank), 3)
print(rank)
# test for prediction while considering only a subset of ratings from the user
small_ratings = pd.DataFrame.from_records(
[("2", "tt0112453", 0.4)], columns=["from_id", "to_id", "score"])
alg = NXPageRank(graph,
item_feature_selection_algorithm=FSPageRank(1),
user_feature_selection_algorithm=FSPageRank(1))
rank = alg.predict(small_ratings, 3)
self.assertEqual(len(rank), 3)
print(rank)
# test for prediction while considering only a subset of ratings from the user with an item not present in the
# original rating frame
wrong_small_ratings = pd.DataFrame.from_records(
[("2", "tt0112453", 0.4), ("2", "test", 0.3)],
columns=["from_id", "to_id", "score"])
alg = NXPageRank(graph,
item_feature_selection_algorithm=FSPageRank(1),
user_feature_selection_algorithm=FSPageRank(1))
rank = alg.predict(wrong_small_ratings, 3)
self.assertEqual(len(rank), 0)
user_ratings = ratings[ratings['from_id'] == '8']
# test for prediction with user ratings containing only the most negative vote as possible
alg = NXPageRank(graph,
item_feature_selection_algorithm=FSPageRank(1),
user_feature_selection_algorithm=FSPageRank(0))
rank = alg.predict(user_ratings, 3)
self.assertEqual(len(rank), 3)
print(rank)