def sindexf(decomp, h):
'''
Projects the seasonal component of a seasonal decomposition of a time series
forward by h time steps into the future.
Args:
decomp: an R seasonal decomposition from stl or decompose,
or a Pandas data frame containing a seasonal decomposition.
h: a forecast horizon
Returns:
an object that maps to am R time series containing the seasonal component
of decomp, projected naively forward h steps.
'''
if validate.is_Pandas_decomposition(decomp):
freq = len(decomp.index.levels[1])
seasonal = list(decomp.seasonal[-freq:])
it = itertools.cycle(seasonal)
data = []
for k in range(h):
data.append(it.next())
data = [0] + data
start = decomp.seasonal.last_valid_index()
series = converters.sequence_as_series(data, start=start, freq=freq)
return series[1:]
elif validate.is_R_decomposition(decomp):
return fc.sindexf(decomp, h)
else:
raise ValueError('seasadj requires a seasonal decomposition as input')
def seasadj(decomp):
'''
Return a seasonally adjusted version of the origin time series that
was seasonally decomposed to get decomp.
Args:
decomp: an R seasonal decomposition from stl or decompose,
or a Pandas data frame containing a seasonal decomposition.
Returns:
an object that maps an R time series of the seasonally adjusted
values of the series that decomp was formed from
'''
if validate.is_Pandas_decomposition(decomp):
if decomp.trend.isnull().any():
# classical
seasonal_mean = decomp.seasonal.mean(axis=0)
if numpy.allclose(seasonal_mean, 1.0, atol=1e-5):
# multiplicative
return decomp.data / decomp.seasonal
else:
# additive
return decomp.data - decomp.seasonal
else:
# stl decomposition
return decomp.data - decomp.seasonal
elif validate.is_R_decomposition(decomp):
return fc.seasadj(decomp)
else:
raise ValueError('seasadj requires a seasonal decomposition as input')
def sindexf(decomp, h):
'''
Projects the seasonal component of a seasonal decomposition of a time series
forward by h time steps into the future.
Args:
decomp: an R seasonal decomposition from stl or decompose,
or a Pandas data frame containing a seasonal decomposition.
h: a forecast horizon
Returns:
an object that maps to am R time series containing the seasonal component
of decomp, projected naively forward h steps.
'''
if validate.is_Pandas_decomposition(decomp):
freq = len(decomp.index.levels[1])
seasonal = list(decomp.seasonal[-freq:])
it = itertools.cycle(seasonal)
data = []
for k in range(h):
data.append(it.next())
data = [0] + data
start = decomp.seasonal.last_valid_index()
series = converters.sequence_as_series(data, start=start, freq=freq)
return series[1:]
elif validate.is_R_decomposition(decomp):
return fc.sindexf(decomp, h)
else:
raise ValueError('seasadj requires a seasonal decomposition as input')
def seasadj(decomp):
'''
Return a seasonally adjusted version of the origin time series that
was seasonally decomposed to get decomp.
Args:
decomp: an R seasonal decomposition from stl or decompose,
or a Pandas data frame containing a seasonal decomposition.
Returns:
an object that maps an R time series of the seasonally adjusted
values of the series that decomp was formed from
'''
if validate.is_Pandas_decomposition(decomp):
if decomp.trend.isnull().any():
# classical
seasonal_mean = decomp.seasonal.mean(axis=0)
if numpy.allclose(seasonal_mean, 1.0, atol=1e-5):
# multiplicative
return decomp.data / decomp.seasonal
else:
# additive
return decomp.data - decomp.seasonal
else:
# stl decomposition
return decomp.data - decomp.seasonal
elif validate.is_R_decomposition(decomp):
return fc.seasadj(decomp)
else:
raise ValueError('seasadj requires a seasonal decomposition as input')
def to_decomp(dc):
'''
Accepts either an R decomposition or a Pandas Data Frame containing a
decomposition. In either case, it returns a Pandas Data Frame.
Args:
dc: an R decomposition or a Pandas Data Frame containing a decomposition
Returns:
the decomposition in dc, as a Pandas Data Frame
'''
if validate.is_Pandas_decomposition(dc):
return dc
elif validate.is_R_decomposition(dc):
return decomposition(dc)
else:
raise TypeError('Must provide a Pandas-type or R-type decomposition')
def to_decomp(dc):
'''
Accepts either an R decomposition or a Pandas Data Frame containing a
decomposition. In either case, it returns a Pandas Data Frame.
Args:
dc: an R decomposition or a Pandas Data Frame containing a decomposition
Returns:
the decomposition in dc, as a Pandas Data Frame
'''
if validate.is_Pandas_decomposition(dc):
return dc
elif validate.is_R_decomposition(dc):
return decomposition(dc)
else:
raise TypeError('Must provide a Pandas-type or R-type decomposition')