def test_fit_1():
"""Test PennAIClassifier fit() without setting knowledgebase/kb_metafeatures."""
pennai = PennAIClassifier(
rec_class=RandomRecommender,
n_recs=2,
n_iters=2,
random_state=42,
verbose=0
)
pennai.fit(X_train, y_train)
assert pennai.score(X_test, y_test)
def test_verbose_0():
"""Test PennAIClassifier fit() with verbose=0."""
pennai = PennAIClassifier(
rec_class=RandomRecommender,
n_recs=2,
n_iters=2,
knowledgebase=classification_kb,
kb_metafeatures=classification_metafeatures,
random_state=42,
verbose=0
)
pennai.fit(X_train, y_train)
assert pennai.score(X_train, y_train)
def test_fit_init():
"""Test PennAIClassifier fit_init_() works as expected."""
pennai = PennAIClassifier(
rec_class=RandomRecommender,
n_recs=2,
n_iters=2,
knowledgebase=classification_kb,
kb_metafeatures=classification_metafeatures,
random_state=42
)
pennai._fit_init()
assert pennai.scoring_ == "accuracy"
assert pennai.metric_ == "accuracy"
assert pennai.n_recs_ == 2
def test_PennAIClassifier_fit(recommender):
"""Test PennAIClassifier fit() with all Recommenders."""
print("\nStart fit test for ", recommender.__name__)
if recommender.__name__ not in ["CoClusteringRecommender"]:
pennai = PennAIClassifier(
rec_class=recommender,
n_recs=2,
n_iters=2,
knowledgebase=classification_kb,
kb_metafeatures=classification_metafeatures,
random_state=42,
verbose=0
)
pennai.fit(X_train, y_train)
assert pennai.score(X_test, y_test)
def test_init():
"""Test PennAIClassifier __init__() can assign correct params"""
pennai = PennAIClassifier(
rec_class=RandomRecommender,
n_recs=2,
n_iters=2,
knowledgebase=classification_kb,
kb_metafeatures=classification_metafeatures,
random_state=42
)
assert pennai.n_recs == 2
assert pennai.n_iters == 2
assert pennai.rec_class == RandomRecommender
assert pennai.knowledgebase == classification_kb
assert pennai.kb_metafeatures == classification_metafeatures
assert pennai.random_state == 42
assert pennai.verbose == 0
assert pennai.serialized_rec == None
assert pennai.scoring == None
assert pennai.config_dict == None
assert pennai.ensemble == None
assert pennai.max_time_mins == None
assert pennai.stopping_criteria == None
assert pennai.mode == "classification"
assert pennai.n_jobs == 1
def test_warm_start(recommender=SVDRecommender):
"""Test PennAIClassifier fit() with warm_start."""
if recommender.__name__ not in ["AverageRecommender", "RandomRecommender", "CoClusteringRecommender"]:
serialized_rec_directory = "data/recommenders/scikitlearn"
serialized_rec_filename = "{}_classifier_accuracy_pmlb.pkl.gz".format(recommender.__name__)
serialized_rec = path.join(serialized_rec_directory, serialized_rec_filename)
pennai = PennAIClassifier(
rec_class=recommender,
n_recs=2,
n_iters=2,
serialized_rec=serialized_rec,
random_state=42,
verbose=0
)
pennai.fit(X_train, y_train)
assert pennai.score(X_train, y_train)
def test_ensemble():
"""Test PennAIClassifier fit() with ensemble=5."""
pennai = PennAIClassifier(
rec_class=RandomRecommender,
n_recs=5,
n_iters=2,
knowledgebase=classification_kb,
kb_metafeatures=classification_metafeatures,
ensemble=5,
random_state=42,
verbose=0
)
pennai.fit(X_train, y_train)
assert pennai.score(X_train, y_train)
assert pennai.ensemble
assert pennai.estimator.__class__ == VotingClassifier
def test_load_kb():
"""Test PennAIClassifier load_kb can load correct kb."""
pennai = PennAIClassifier(
rec_class=RandomRecommender,
n_recs=2,
n_iters=2,
random_state=42
)
pennai.dataset_mf_cache = pd.DataFrame()
pennai.kb_ = classification_kb
pennai.mf_ = classification_metafeatures
resultsData = pennai._load_kb()
kb = load_knowledgebase(
resultsFiles=[classification_kb],
metafeaturesFiles=[classification_metafeatures]
)
all_df_mf = kb['metafeaturesData'].set_index('_id', drop=False)
df = all_df_mf.loc[kb['resultsData']["classification"]['_id'].unique()]
assert pennai.dataset_mf_cache.equals(df)
assert kb['resultsData'][pennai.mode].equals(resultsData)