def test_add_time_day():
# Creating hourly time stamp for three days
# 2017-09-16 ==> Saturday
# 2017-09-17 ==> Sunday
# 2017-09-18 ==> Monday
date_hr_timestamp = pd.date_range('2017-09-16', periods=72, freq='H', tz=pytz.UTC)
df = pd.DataFrame( {'energy' : [1.0 for xx in date_hr_timestamp]} , index=date_hr_timestamp)
day_of_week = HourlyDayOfWeekModel()
returned_df = day_of_week.add_time_day(df)
assert 'hour_of_day' in returned_df
assert 'day_of_week' in returned_df
# Testing day of week columns
# 2017-09-16 is Saturday and so day of week value of the first row
# in returned_df should be 5
assert returned_df.get_value(returned_df.index[0], 'day_of_week') == '5'
#2017-09-19 is Monday and so day of week value of last row should 0
assert returned_df.get_value(returned_df.index[-1], 'day_of_week') == '0'
# 2017-09-18 is Sunday, day_of_week should be 6
assert returned_df.get_value(returned_df.index[25], 'day_of_week') == '6'
# First hour of 2017-09-16
assert returned_df.get_value(returned_df.index[1], 'hour_of_day') == '1'
# Second hour of 2017-09-16
assert returned_df.get_value(returned_df.index[2], 'hour_of_day') == '2'
def test_predict(input_df):
model = HourlyDayOfWeekModel(min_contiguous_months=0)
model.fit(input_df)
# Test cases on the output data types of predict function.
# When summed = True, prediction and variance are atomic
# floating point numbers
prediction, variance= model.predict(input_df, summed=True)
assert type(prediction) == np.float64
#assert type(variance) == np.float64
# When summed=False, prediction and variance are pandas Series.
prediction, variance= model.predict(input_df, summed=False)
assert type(prediction) == pd.Series
assert type(variance) == pd.Series
# We expect out linear regression model to make prediction between 0.95 & 2.0 because the energy
# consumed in input_df dataframe is always 1.0, please take a look at hourly_trace_with_dummy_energy function
assert prediction[0].item() > 0.95 and prediction[0].item() < 2.0
def test_min_contiguous_months(input_df):
min_contiguous_months = 9
model = HourlyDayOfWeekModel(min_contiguous_months=min_contiguous_months)
with pytest.raises(model_exceptions.DataSufficiencyException) as sufficiency_exception:
model.fit(input_df)
def test_add_cdd(input_df):
cdd_function = HourlyDayOfWeekModel()
cdd_val= cdd_function.add_cdd(input_df)
assert 'cdd' in cdd_val
def test_add_hdd(input_df):
hdd_function = HourlyDayOfWeekModel()
hdd_val= hdd_function.add_hdd(input_df)
assert 'hdd' in hdd_val