def test_extract_features_with_wrong_feature_list():
HERE = pathlib.Path(__file__).parents[3]
test_df = utils.read_to_df(HERE / 'source/CyberBullying/dataNew.csv')
with pytest.raises(ValueError):
feature_list = ['wrong', 'feature', 'list']
fe.extract_features(test_df, feature_list)
def test_extract_features_with_bad_file():
with pytest.raises(ValueError):
test_df = pd.DataFrame(['not valid'], columns=["unknown"])
feature_list = [
'post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis',
'not_off_dis'
]
fe.extract_features(test_df, feature_list)
def test_feature_extraction():
try:
tagged_df = utils.read_to_df(ROOT + '/source/CyberBullying/dataNew.csv')
pre.preprocess(tagged_df)
assert isinstance(fe.extract_features(tagged_df, FEATURE_LIST), pd.DataFrame)
except Exception:
pytest.fail('Unexpected error..')
def test_correct_features():
raw_df = pd.DataFrame(
[[1, 'מילה עוד מילה חמודה', 0], [2, 'את מטומטמת', 3]],
columns=['id', 'text', 'cb_level'])
feature_list = [
'post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis',
'not_off_dis'
]
test_df = fe.extract_features(raw_df, feature_list)
expected_columns = [
'id', 'post_length', 'tfidf', 'T1', 'T2', 'T3', 'screamer', 'off_dis',
'not_off_dis'
]
if test_df.isnull().values.any():
assert False
for col in expected_columns:
assert col in test_df.columns
assert test_df.iloc[0]['post_length'] == 4
assert test_df.iloc[1]['post_length'] == 2
assert test_df.iloc[0]['tfidf'] > 0
assert test_df.iloc[1]['tfidf'] > 0
assert test_df.iloc[0]['off_dis'] > 0
assert test_df.iloc[1]['off_dis'] > 0
assert test_df.iloc[0]['not_off_dis'] > 0
assert test_df.iloc[1]['not_off_dis'] > 0
assert test_df.iloc[0]['T1'] > 0
assert test_df.iloc[1]['T1'] > 0
assert test_df.iloc[0]['T2'] > 0
assert test_df.iloc[1]['T2'] > 0
assert test_df.iloc[0]['T3'] > 0
assert test_df.iloc[1]['T3'] > 0
def test_train_model():
rf_obj = rf.RandomForest()
tagged_df = utils.read_to_df(ROOT + '/source/CyberBullying/dataNew.csv')
tagged_df = pre.preprocess(tagged_df)
feature_list = ['post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis', 'not_off_dis']
X = fe.extract_features(tagged_df, feature_list)
X = X.drop(columns=['id'])
y = (tagged_df['cb_level'] == 3).astype(int)
rf_obj.train(X, y)
assert rf_obj.model is not None
def test_explain_class():
model = utils.get_model(ROOT + '/source/CyberBullying/outputs/RandomForest.pkl')
post = 'אני אוהבת אותך'
tagged_df = pd.DataFrame({'id': [1], 'text': [post]})
tagged_df = pre.preprocess(tagged_df)
feature_list = ['post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis', 'not_off_dis']
X = fe.extract_features(tagged_df, feature_list)
X = X.drop(columns=['id'])
explanation.explain_class(model, X)
assert os.path.isfile(ROOT + '/source/CyberBullying/outputs/force_plot_post.png')
def test_explain_dataset():
model = utils.get_model(ROOT + '/source/CyberBullying/outputs/RandomForest.pkl')
tagged_df = utils.read_to_df(ROOT + '/source/CyberBullying/dataNew.csv')
tagged_df = pre.preprocess(tagged_df)
feature_list = ['post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis', 'not_off_dis']
X = fe.extract_features(tagged_df, feature_list)
X = X.drop(columns=['id'])
explanation.explain_model(model, X)
assert os.path.isfile(ROOT + '/source/CyberBullying/outputs/dependence_plot.png')
assert os.path.isfile(ROOT + '/source/CyberBullying/outputs/summary_plot_bar.png')
assert os.path.isfile(ROOT + '/source/CyberBullying/outputs/summary_plot.png')
def test_explainability():
try:
model = utils.get_model(ROOT + '/source/CyberBullying/outputs/RandomForest.pkl')
tagged_df = utils.read_to_df(ROOT + '/source/CyberBullying/dataNew.csv')
tagged_df = pre.preprocess(tagged_df)
X = fe.extract_features(tagged_df, FEATURE_LIST)
X = X.drop(columns=['id'])
explanation.explain_model(model, X)
explanation.explain_class(model, X)
except Exception:
pytest.fail('Unexpected error..')
def test_classicifation():
try:
tagged_df = utils.read_to_df(ROOT + '/source/CyberBullying/dataNew.csv')
pre.preprocess(tagged_df)
X = fe.extract_features(tagged_df, FEATURE_LIST)
y = (tagged_df['cb_level'] == 3).astype(int)
X = X.drop(columns=['id'])
rf_obj = rf.RandomForest()
rf_obj.train(X, y)
rf_obj.predict(X)
assert os.path.isfile(ROOT + '/source/CyberBullying/outputs/RandomForest.pkl')
except Exception:
pytest.fail('Unexpected error..')
def test_correct_classification():
model = utils.get_model(ROOT + '/source/CyberBullying/outputs/RandomForest.pkl')
rf_obj = rf.RandomForest()
rf_obj.model = model
post = 'אני אוהבת אותך'
tagged_df = pd.DataFrame({'id': [1], 'text': [post]})
tagged_df = pre.preprocess(tagged_df)
feature_list = ['post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis', 'not_off_dis']
X = fe.extract_features(tagged_df, feature_list)
X = X.drop(columns=['id'])
y_prob_rf = rf_obj.predict(X)
my_prob = y_prob_rf[0]
if my_prob is None:
assert False
assert isinstance(my_prob, numpy.float64)
def test_performance():
model = utils.get_model(ROOT + '/source/CyberBullying/outputs/RandomForest.pkl')
rf_obj = rf.RandomForest()
rf_obj.model = model
df = utils.read_to_df(ROOT + '/source/CyberBullying/dataNew.csv')
df = pre.preprocess(df)
X = fe.extract_features(df, FEATURE_LIST)
X = X.drop(columns=['id'])
y = (df['cb_level'] == 3).astype(int)
y_prob_rf = rf_obj.predict(X)
pred = np.where(y_prob_rf > 0.5, 1, 0)
performance = per.get_performances(y, pred)
assert 'f-score' in performance.keys()
assert 'recall' in performance.keys()
assert 'precision' in performance.keys()
get tagged df
"""
tagged_df = utils.read_to_df() # Vigo data
# tagged_df = utils.create_csv_from_keepers_files() # Keepers data
"""
Run a pre-processing function on the tagged data
"""
tagged_df = pre.preprocess(tagged_df)
"""
Run a extract features function on the clean and tagged data
"""
feature_list = ['post_length', 'tfidf', 'topics', 'screamer', 'words', 'off_dis', 'not_off_dis']
fe.folder_name = logger.folder_name
X = fe.extract_features(tagged_df, feature_list)
logger.write_features(feature_list)
y = (tagged_df['cb_level'] == 3).astype(int)
X = X.drop(columns=['id'])
"""
Split data to train and test
"""
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
performances_list = {}
auc_list = {}
"""
Running the baseline model
"""
