def test_two_categories_high_min_value():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy4.txt')
df = read_data(path)
twitter = TwitterAnomalyTrendinessDetector(freq='12H',
is_multicategory=False)
prediction = twitter.predict(df)
assert len(prediction[prediction['prediction'] == 1]) == 0
def test_single_category():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy4.txt')
df = read_data(path)
one_category = df.loc[pd.IndexSlice[:, 'housing'], :].reset_index(level='category', drop=True)
ma = STLTrendinessDetector(freq='12H', is_multicategory=False, anomaly_type='or')
prediction = ma.predict(one_category)
assert len(prediction[prediction['prediction'] == 1]) == 1
def test_read_data():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy.txt')
df = read_data(path)
index = df.index
# assert index is pd.MultiIndex
assert len(df) == 16
assert isinstance(df.index, pd.MultiIndex)
def test_two_categories():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy4.txt')
df = read_data(path)
ma = MovingAverageSeasonalTrendinessDetector(freq='12H',
anomaly_type='or',
lookback='5D')
prediction = ma.predict(df)
assert len(prediction[prediction['prediction'] == 1]) == 2
def test_two_categories_high_min_value():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy4.txt')
df = read_data(path)
ma = STLTrendinessDetector(freq='12H',
anomaly_type='or',
lookback='5D',
min_value=2000)
prediction = ma.predict(df)
assert len(prediction[prediction['prediction'] == 1]) == 0
def test_y_None():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy.txt')
df = read_data(path)
X = df[['value']]
y = None
model = MockModel()
models = [model]
with pytest.raises(TypeError):
eval_models(X, y, models, label_col_name='is_anomaly')
def test_real_model():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy.txt')
df = read_data(path)
model = MovingAverageSeasonalTrendinessDetector(is_multicategory=True,
freq='12H')
models = [model]
X = df[['value']]
y = df[['is_anomaly']]
res = eval_models(X, y, models, label_col_name='is_anomaly')
print(res)
def test_eval_models_all_false():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy.txt')
df = read_data(path)
df['is_anomaly'] = 0
model = MockModel()
models = [model]
X = df[['value']]
y = df[['is_anomaly']]
res = eval_models(X, y, models, label_col_name='is_anomaly')
assert math.isnan(res['MockModel']['f1'])
assert res['MockModel']['precision'] == 0
assert math.isnan(res['MockModel']['recall'])
def test_n_splits_big():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy2.txt')
df = read_data(path)
model = MockModel()
models = [model]
X = df[['value']]
y = df[['is_anomaly']]
res = eval_models_CV(X,
y,
models,
n_splits=40000,
verbose=True,
label_col_name='is_anomaly')
assert res['MockModel']['f1'] == 1.0
assert res['MockModel']['precision'] == 1.0
assert res['MockModel']['recall'] == 1.0
def test_eval_models_half_false():
path = os.path.join(THIS_DIR, os.pardir, 'tests/data/dummy2.txt')
df = read_data(path)
df['is_anomaly'] = 0
df.iloc[-1]['is_anomaly'] = 1
df.iloc[-2]['is_anomaly'] = 1
model = MockModel()
models = [model]
X = df[['value']]
y = df[['is_anomaly']]
res = eval_models(X,
y,
models,
label_col_name='is_anomaly',
window_size_for_metrics=0)
assert res['MockModel']['precision'] == 0.5
assert res['MockModel']['recall'] == 1.0
