def test_perform_no_zeros(self):
# Save only 'no-zeros' graph
metric = PopRecsCorrelation('test_pop_recs/no_zeros', mode='no_zeros')
metric.perform(split_i21_missing_in_recs)
self.assertFalse(os.path.isfile(os.path.join('test_pop_recs/no_zeros', 'pop_recs_correlation.png')))
# The 'no-zeros' graph is created adding with file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/no_zeros', 'pop_recs_correlation_no_zeros.png')))
def test_perform_w_zeros(self):
# Save only 'w-zeros' graph
metric = PopRecsCorrelation('test_pop_recs/w_zeros', mode='w_zeros')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/w_zeros', 'pop_recs_correlation.png')))
# If The 'no-zeros' graph was created, its file_name would be: file_name = file_name + '_no_zeros'
self.assertFalse(os.path.isfile(os.path.join('test_pop_recs/w_zeros', 'pop_recs_correlation_no_zeros.png')))
def test_perform_both(self):
# Save both graph when it makes sense (eg. they are different)
metric = PopRecsCorrelation('test_pop_recs/both_yes', mode='both')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/both_yes', 'pop_recs_correlation.png')))
# If The 'no-zeros' is created, its file_name will be: file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/both_yes', 'pop_recs_correlation_no_zeros.png')))
truth = pd.DataFrame({
'from_id': ['u1', 'u1', 'u2', 'u2', 'u2'],
'to_id': ['i2', 'i1', 'i3', 'i5', 'i4'],
'score': [5, 3, 3.6, 4, 2.2]}
)
recs = pd.DataFrame({
'from_id': ['u1', 'u1', 'u1', 'u2', 'u2', 'u2', 'u2'],
'to_id': ['i1', 'i2', 'inew1', 'inew2', 'i5', 'i4', 'i3'],
'score': [300, 250, 200, 400, 350, 300, 100]}
)
# All items in the truth set have been recommended, so there's no 'zero' recommendation
split_no_zero_present = Split(recs, truth)
metric = PopRecsCorrelation('test_pop_recs/both_identical', mode='both')
metric.perform(split_no_zero_present)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/both_identical', 'pop_recs_correlation.png')))
# If The 'no-zeros' is created, its file_name will be: file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/both_identical', 'pop_recs_correlation_no_zeros.png')))
def test_graph(self):
catalog = set(ratings.to_id)
users_dir = os.path.join(dir_test_files, 'complex_contents',
'users_codified/')
graph = NXFullGraph(
ratings,
user_contents_dir=users_dir,
item_contents_dir=items_dir,
item_exo_representation="dbpedia",
user_exo_representation='local',
item_exo_properties=['starring'],
user_exo_properties=['1'
] # It's the column in the users .DAT which
# identifies the gender
)
graph_rs = GraphBasedRS(NXPageRank(), graph)
em = EvalModel(graph_rs,
KFoldPartitioning(),
metric_list=[
Precision(relevant_threshold=3),
Recall(),
FMeasure(beta=1),
FMeasure(beta=2, sys_average='micro'),
MRR(),
Correlation('pearson'),
GiniIndex(),
DeltaGap({
'popular': 0.5,
'niche': 0.5
}),
PredictionCoverage(catalog),
PopProfileVsRecs(user_groups={
'popular': 0.5,
'niche': 0.5
},
out_dir='plots/'),
LongTailDistr('plots/', format='svg'),
PopRecsCorrelation('plots/')
],
verbose_predictions=True,
methodology=TestItemsMethodology())
em.fit()
def test_eval_ranking_needed_metrics_implicit_split(self):
# We set the split_list directly by the class attribute
c = MetricCalculator()
RankingNeededMetric.rank_truth_list = self.rank_split_list
system_res, each_user_res = c.eval_metrics([
Precision(),
PrecisionAtK(2),
RPrecision(),
Recall(),
RecallAtK(2),
FMeasure(),
FMeasureAtK(2),
NDCG(),
NDCGAtK(2),
MRR(),
MRRAtK(2),
Correlation('pearson'),
Correlation('kendall'),
Correlation('spearman'),
PredictionCoverage(self.catalog),
CatalogCoverage(self.catalog, top_n=2),
GiniIndex(),
DeltaGap(user_groups={
'a': 0.5,
'b': 0.5
}),
LongTailDistr(out_dir='test_plot'),
PopProfileVsRecs(user_groups={
'a': 0.5,
'b': 0.5
},
out_dir='test_plot'),
PopRecsCorrelation(out_dir='test_plot')
])
self.assertIsInstance(system_res, pd.DataFrame)
self.assertIsInstance(each_user_res, pd.DataFrame)
def test_overwrite(self):
# Save both graph when it makes sense (eg. they are different)
metric = PopRecsCorrelation('test_pop_recs/overwrite', mode='both')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation.png')))
# If The 'no-zeros' is created, its file_name will be: file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation_no_zeros.png')))
# Save both graph already existent for the first time
metric = PopRecsCorrelation('test_pop_recs/overwrite', mode='both')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation (1).png')))
# If The 'no-zeros' is created, its file_name will be: file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation_no_zeros (1).png')))
# Save both graph already existent for the second time
metric = PopRecsCorrelation('test_pop_recs/overwrite', mode='both')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation (2).png')))
# If The 'no-zeros' is created, its file_name will be: file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation_no_zeros (2).png')))
# Save both graph already existent but overwrite them
metric = PopRecsCorrelation('test_pop_recs/overwrite', mode='both', overwrite=True)
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation.png')))
# If The 'no-zeros' is created, its file_name will be: file_name = file_name + '_no_zeros'
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation_no_zeros.png')))
self.assertFalse(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation (3).png')))
self.assertFalse(os.path.isfile(os.path.join('test_pop_recs/overwrite', 'pop_recs_correlation (3).png')))
def test_perform(self):
# Save on same folder
metric = PopRecsCorrelation()
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile('./pop_recs_correlation.png'))
self.assertTrue(os.path.isfile('./pop_recs_correlation_no_zeros.png'))
# Save on same folder with a different format (svg)
metric = PopRecsCorrelation(format='svg')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile('./pop_recs_correlation.svg'))
self.assertTrue(os.path.isfile('./pop_recs_correlation_no_zeros.svg'))
# Save on a not existent folder with a specified file_name
metric = PopRecsCorrelation('test_pop_recs', file_name='pop_recs_custom_name')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs', 'pop_recs_custom_name.png')))
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs', 'pop_recs_custom_name_no_zeros.png')))
# Save on an existent folder with a specified file_name and with a different format (svg)
# Save also the frame used to build the box_plot
metric = PopRecsCorrelation('test_pop_recs', file_name='pop_recs_custom_name', format='svg')
metric.perform(split_i21_missing_in_recs)
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs', 'pop_recs_custom_name.svg')))
self.assertTrue(os.path.isfile(os.path.join('test_pop_recs', 'pop_recs_custom_name_no_zeros.svg')))