def test_time_diff_in_quarters():
""" Test convert time difference to nearest quarters: """
test_data = [['2005-01', '2005-04', 1],
['2015-01-01', '2015-08-07', 2],
['2015-01-01', '2015-07-07', 2], # Check that it rounds up.
['2015-12-31', '2015-01-01', -4], # Check negative.
]
for [from_time_str, to_time_str, d_quarter] in test_data:
t1 = np.datetime64(from_time_str)
t2 = np.datetime64(to_time_str)
dt = time_tools.time_diff_in_quarters(t1, t2)
nose.tools.eq_(dt, d_quarter, ''.join((
'Error in delta quarter result, expected:',
str(d_quarter), 'return was:', str(dt))))
def test_time_diff_in_quarters():
""" Test convert time difference to nearest quarters: """
test_data = [
['2005-01', '2005-04', 1],
['2015-01-01', '2015-08-07', 2],
['2015-01-01', '2015-07-07', 2], # Check that it rounds up.
['2015-12-31', '2015-01-01', -4], # Check negative.
]
for [from_time_str, to_time_str, d_quarter] in test_data:
t1 = np.datetime64(from_time_str)
t2 = np.datetime64(to_time_str)
dt = time_tools.time_diff_in_quarters(t1, t2)
nose.tools.eq_(
dt, d_quarter, ''.join(('Error in delta quarter result, expected:',
str(d_quarter), 'return was:', str(dt))))
def remap_to_relative_time(self,
prediction_data,
actual_data,
prediction_horizon=16):
""" Convert the data to columns of how old the prediction is.
"""
# Create a new empty DataFrame, and add the CPI column.
index = self.raw_data.index
slength = len(index)
pred_relative = pd.DataFrame(index=index, data=None)
colname_prefix = actual_data.name
# Create the column for relative predictions, and fill with NaN.
pred_relative_col_names = []
# TODO(Camilla): Hvis språkbruken er "ett kvartal frem" for første
# prediksjon i banen, bytt om til range(1, prediction_horizon+1).
for dq in range(0, prediction_horizon):
col_name = colname_prefix + '_dQ' + str(dq)
pred_relative_col_names.append(col_name)
ser = pd.Series(np.empty(slength), index=index)
pred_relative[col_name] = ser
pred_relative[col_name] = np.NaN
# Okay now we have an empty frame. Let's fill it up...
# Loop through all the PPR columns in CPI predictions.
for col_name in prediction_data:
col_info = decode_column_name.decode(col_name)
ppr_date = time_tools.year_quarter_to_datetime(
col_info['year'], col_info['quarter'])
# Loop through each prediction in this PPR.
for date in prediction_data.index:
if date is pd.NaT:
continue
prediction = prediction_data.loc[date, col_name]
if not math.isnan(prediction):
# How old is the prediction, in quarters?
dq = time_tools.time_diff_in_quarters(ppr_date, date)
n_col_name = pred_relative_col_names[dq]
# Insert prediction, in the new dataframe.
pred_relative.loc[date, n_col_name] = (
prediction_data.loc[date, col_name])
#print 'Put prediction from ', col_name,
# ' for ', date.strftime('%Y-%m-%d'),
# ' in ', n_col_name
return pred_relative
def remap_to_relative_time(self, prediction_data, actual_data,
prediction_horizon=16):
""" Convert the data to columns of how old the prediction is.
"""
# Create a new empty DataFrame, and add the CPI column.
index = self.raw_data.index
slength = len(index)
pred_relative = pd.DataFrame(index=index, data=None)
colname_prefix = actual_data.name
# Create the column for relative predictions, and fill with NaN.
pred_relative_col_names = []
# TODO(Camilla): Hvis språkbruken er "ett kvartal frem" for første
# prediksjon i banen, bytt om til range(1, prediction_horizon+1).
for dq in range(0, prediction_horizon):
col_name = colname_prefix + '_dQ' + str(dq)
pred_relative_col_names.append(col_name)
ser = pd.Series(np.empty(slength), index=index)
pred_relative[col_name] = ser
pred_relative[col_name] = np.NaN
# Okay now we have an empty frame. Let's fill it up...
# Loop through all the PPR columns in CPI predictions.
for col_name in prediction_data:
col_info = decode_column_name.decode(col_name)
ppr_date = time_tools.year_quarter_to_datetime(
col_info['year'], col_info['quarter'])
# Loop through each prediction in this PPR.
for date in prediction_data.index:
if date is pd.NaT:
continue
prediction = prediction_data.loc[date, col_name]
if not math.isnan(prediction):
# How old is the prediction, in quarters?
dq = time_tools.time_diff_in_quarters(ppr_date, date)
n_col_name = pred_relative_col_names[dq]
# Insert prediction, in the new dataframe.
pred_relative.loc[date, n_col_name] = (
prediction_data.loc[date, col_name])
#print 'Put prediction from ', col_name,
# ' for ', date.strftime('%Y-%m-%d'),
# ' in ', n_col_name
return pred_relative