Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Factorizing reverse_sigmah from HierarchicalReconciliation #121

Merged
merged 2 commits into from
Dec 6, 2022
Merged

Factorizing reverse_sigmah from HierarchicalReconciliation #121

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
33d5780
factorizing reverse sigmah from core
kdgutier Dec 6, 2022
ac70952
Merge branch 'main' into reverse_engineer_sigmah
kdgutier Dec 6, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion hierarchicalforecast/_modidx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,9 @@
'hierarchicalforecast.core.HierarchicalReconciliation.reconcile': ( 'core.html#hierarchicalreconciliation.reconcile',
'hierarchicalforecast/core.py'),
'hierarchicalforecast.core._build_fn_name': ( 'core.html#_build_fn_name',
'hierarchicalforecast/core.py')},
'hierarchicalforecast/core.py'),
'hierarchicalforecast.core._reverse_engineer_sigmah': ( 'core.html#_reverse_engineer_sigmah',
'hierarchicalforecast/core.py')},
'hierarchicalforecast.evaluation': { 'hierarchicalforecast.evaluation.HierarchicalEvaluation': ( 'evaluation.html#hierarchicalevaluation',
'hierarchicalforecast/evaluation.py'),
'hierarchicalforecast.evaluation.HierarchicalEvaluation.__init__': ( 'evaluation.html#hierarchicalevaluation.__init__',
Expand Down
53 changes: 37 additions & 16 deletions hierarchicalforecast/core.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,35 @@ def _build_fn_name(fn) -> str:
return fn_name

# %% ../nbs/core.ipynb 9
def _reverse_engineer_sigmah(Y_hat_df, y_hat_model, model_name, uids):
"""
This function assumes that the model creates prediction intervals
under a normality assumption with the following the Equation:
$\hat{y}_{t+h} + c \hat{sigma}_{h}$

In the future, we might deprecate this function in favor of a
direct usage of an estimated $\hat{sigma}_{h}$
"""
drop_cols = ['ds', 'y'] if 'y' in Y_hat_df.columns else ['ds']
model_names = Y_hat_df.drop(columns=drop_cols, axis=1).columns.to_list()
pi_model_names = [name for name in model_names if ('-lo' in name or '-hi' in name)]
pi_model_name = [pi_name for pi_name in pi_model_names if model_name in pi_name]
pi = len(pi_model_name) > 0

if not pi:
raise Exception(f'Please include {model_name} prediction intervals in `Y_hat_df`')

pi_col = pi_model_name[0]
sign = -1 if 'lo' in pi_col else 1
level_col = re.findall('[\d]+[.,\d]+|[\d]*[.][\d]+|[\d]+', pi_col)
level_col = float(level_col[0])
z = norm.ppf(0.5 + level_col / 200)
sigmah = Y_hat_df.pivot(columns='ds', values=pi_col).loc[uids].values
sigmah = sign * (sigmah - y_hat_model) / z

return sigmah

# %% ../nbs/core.ipynb 10
class HierarchicalReconciliation:
"""Hierarchical Reconciliation Class.

Expand Down Expand Up @@ -138,24 +167,16 @@ def reconcile(self,
reconcile_fn_name = _build_fn_name(reconcile_fn)
has_fitted = 'y_hat_insample' in signature(reconcile_fn).parameters
has_level = 'level' in signature(reconcile_fn).parameters

for model_name in model_names:
# should we calculate prediction intervals?
pi_model_name = [pi_name for pi_name in pi_model_names if model_name in pi_name]
pi = len(pi_model_name) > 0
# Remember: pivot sorts uid
y_hat_model = Y_hat_df.pivot(columns='ds', values=model_name).loc[uids].values
if pi and has_level and level is not None and intervals_method in ['normality', 'permbu']:
# we need to construct sigmah and add it
# to the reconciler_args
# to recover sigmah we only need
# one prediction intervals
pi_col = pi_model_name[0]
sign = -1 if 'lo' in pi_col else 1
level_col = re.findall('[\d]+[.,\d]+|[\d]*[.][\d]+|[\d]+', pi_col)
level_col = float(level_col[0])
z = norm.ppf(0.5 + level_col / 200)
sigmah = Y_hat_df.pivot(columns='ds', values=pi_col).loc[uids].values
sigmah = sign * (sigmah - y_hat_model) / z

# Recover sigmah and add it to reconciler_args
if has_level and level is not None and intervals_method in ['normality', 'permbu']:
sigmah = _reverse_engineer_sigmah(Y_hat_df=Y_hat_df,
y_hat_model=y_hat_model, model_name=model_name, uids=uids)

reconciler_args['level'] = level
if intervals_method == 'permbu':
y_hat_insample = Y_df.pivot(columns='ds', values=model_name).loc[uids].values
Expand Down Expand Up @@ -202,7 +223,7 @@ def reconcile(self,
# this will require outputs of reconcile_fn to include P, W

fcsts[f'{model_name}/{reconcile_fn_name}'] = fcsts_model['mean'].flatten()
if (pi and has_level and level is not None) or (intervals_method in ['bootstrap'] and level is not None):
if intervals_method in ['bootstrap', 'normality', 'permbu'] and level is not None:
for lv in level:
fcsts[f'{model_name}/{reconcile_fn_name}-lo-{lv}'] = fcsts_model[f'lo-{lv}'].flatten()
fcsts[f'{model_name}/{reconcile_fn_name}-hi-{lv}'] = fcsts_model[f'hi-{lv}'].flatten()
Expand Down
2 changes: 1 addition & 1 deletion hierarchicalforecast/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -244,7 +244,7 @@ def plot_hierarchical_predictions_gap(self,
fig, ax = plt.subplots(figsize=(8, 5))

if 'y' in Y_df.columns:
idx_top = S.sum(axis=1).idxmax()
idx_top = self.S.sum(axis=1).idxmax()
y_plot = Y_df.loc[idx_top].y.values
plt.plot(horizon_dates, y_plot, label='True')

Expand Down
60 changes: 44 additions & 16 deletions nbs/core.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,6 +121,42 @@
"# <span style=\"color:DarkBlue\"> core.HierarchicalReconciliation </span>"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#| exporti\n",
"def _reverse_engineer_sigmah(Y_hat_df, y_hat_model, model_name, uids):\n",
" \"\"\"\n",
" This function assumes that the model creates prediction intervals\n",
" under a normality assumption with the following the Equation:\n",
" $\\hat{y}_{t+h} + c \\hat{sigma}_{h}$\n",
"\n",
" In the future, we might deprecate this function in favor of a \n",
" direct usage of an estimated $\\hat{sigma}_{h}$\n",
" \"\"\"\n",
" drop_cols = ['ds', 'y'] if 'y' in Y_hat_df.columns else ['ds']\n",
" model_names = Y_hat_df.drop(columns=drop_cols, axis=1).columns.to_list()\n",
" pi_model_names = [name for name in model_names if ('-lo' in name or '-hi' in name)]\n",
" pi_model_name = [pi_name for pi_name in pi_model_names if model_name in pi_name]\n",
" pi = len(pi_model_name) > 0\n",
"\n",
" if not pi:\n",
" raise Exception(f'Please include {model_name} prediction intervals in `Y_hat_df`')\n",
"\n",
" pi_col = pi_model_name[0]\n",
" sign = -1 if 'lo' in pi_col else 1\n",
" level_col = re.findall('[\\d]+[.,\\d]+|[\\d]*[.][\\d]+|[\\d]+', pi_col)\n",
" level_col = float(level_col[0])\n",
" z = norm.ppf(0.5 + level_col / 200)\n",
" sigmah = Y_hat_df.pivot(columns='ds', values=pi_col).loc[uids].values\n",
" sigmah = sign * (sigmah - y_hat_model) / z\n",
"\n",
" return sigmah"
]
},
{
"cell_type": "code",
"execution_count": null,
Expand Down Expand Up @@ -231,24 +267,16 @@
" reconcile_fn_name = _build_fn_name(reconcile_fn)\n",
" has_fitted = 'y_hat_insample' in signature(reconcile_fn).parameters\n",
" has_level = 'level' in signature(reconcile_fn).parameters\n",
" \n",
" for model_name in model_names:\n",
" # should we calculate prediction intervals?\n",
" pi_model_name = [pi_name for pi_name in pi_model_names if model_name in pi_name]\n",
" pi = len(pi_model_name) > 0\n",
" # Remember: pivot sorts uid\n",
" y_hat_model = Y_hat_df.pivot(columns='ds', values=model_name).loc[uids].values\n",
" if pi and has_level and level is not None and intervals_method in ['normality', 'permbu']:\n",
" # we need to construct sigmah and add it\n",
" # to the reconciler_args\n",
" # to recover sigmah we only need \n",
" # one prediction intervals\n",
" pi_col = pi_model_name[0]\n",
" sign = -1 if 'lo' in pi_col else 1\n",
" level_col = re.findall('[\\d]+[.,\\d]+|[\\d]*[.][\\d]+|[\\d]+', pi_col)\n",
" level_col = float(level_col[0])\n",
" z = norm.ppf(0.5 + level_col / 200)\n",
" sigmah = Y_hat_df.pivot(columns='ds', values=pi_col).loc[uids].values\n",
" sigmah = sign * (sigmah - y_hat_model) / z\n",
"\n",
" # Recover sigmah and add it to reconciler_args\n",
" if has_level and level is not None and intervals_method in ['normality', 'permbu']:\n",
" sigmah = _reverse_engineer_sigmah(Y_hat_df=Y_hat_df,\n",
" y_hat_model=y_hat_model, model_name=model_name, uids=uids)\n",
"\n",
" reconciler_args['level'] = level\n",
" if intervals_method == 'permbu':\n",
" y_hat_insample = Y_df.pivot(columns='ds', values=model_name).loc[uids].values\n",
Expand Down Expand Up @@ -295,7 +323,7 @@
" # this will require outputs of reconcile_fn to include P, W\n",
"\n",
" fcsts[f'{model_name}/{reconcile_fn_name}'] = fcsts_model['mean'].flatten()\n",
" if (pi and has_level and level is not None) or (intervals_method in ['bootstrap'] and level is not None):\n",
" if intervals_method in ['bootstrap', 'normality', 'permbu'] and level is not None:\n",
" for lv in level:\n",
" fcsts[f'{model_name}/{reconcile_fn_name}-lo-{lv}'] = fcsts_model[f'lo-{lv}'].flatten()\n",
" fcsts[f'{model_name}/{reconcile_fn_name}-hi-{lv}'] = fcsts_model[f'hi-{lv}'].flatten()\n",
Expand Down
6 changes: 3 additions & 3 deletions nbs/utils.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -359,7 +359,7 @@
" fig, ax = plt.subplots(figsize=(8, 5))\n",
" \n",
" if 'y' in Y_df.columns:\n",
" idx_top = S.sum(axis=1).idxmax()\n",
" idx_top = self.S.sum(axis=1).idxmax()\n",
" y_plot = Y_df.loc[idx_top].y.values\n",
" plt.plot(horizon_dates, y_plot, label='True')\n",
"\n",
Expand Down Expand Up @@ -577,9 +577,9 @@
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"display_name": "hierarchicalforecast",
"language": "python",
"name": "python3"
"name": "hierarchicalforecast"
}
},
"nbformat": 4,
Expand Down