Skip to content

Commit

Permalink
choose block size for block bootstrap Pt.3
Browse files Browse the repository at this point in the history
  • Loading branch information
@thierrymoudiki
thierrymoudiki committed Jul 11, 2024
1 parent f16b402 commit 0f2984d
Show file tree
Hide file tree
Showing 7 changed files with 104 additions and 48 deletions.
3 changes: 2 additions & 1 deletion CHANGES.md
View file
Original file line number Diff line number Diff line change
@@ -1,9 +1,10 @@
# version 0.22.4
# version 0.22.7

- Implement new types of predictive simulation intervals (`type_pi`s): independent bootstrap, block bootstrap, 2 variants of split conformal prediction in class `MTS` (see updated docs)
- Gaussian prediction intervals `type_pi == "gaussian"` in class `MTS`
- Implement Winkler score in `LazyMTS` and `LazyDeepMTS` for probabilistic forecasts
- Use conformalized `Estimator`s in `MTS` (see `examples/mts_conformal_not_sims.py`)
- Include `block_size` for block bootstrapping methods for `*MTS` classes

# version 0.20.6

Expand Down
19 changes: 19 additions & 0 deletions nnetsauce/deep/deepMTS.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,23 @@ class DeepMTS(MTS):
lags: int.
number of lags used for each time series.
type_pi: str.
type of prediction interval; currently:
- "gaussian": simple, fast, but: assumes stationarity of Gaussian in-sample residuals and independence in the multivariate case
- "kde": based on Kernel Density Estimation of in-sample residuals
- "bootstrap": based on independent bootstrap of in-sample residuals
- "block-bootstrap": based on basic block bootstrap of in-sample residuals
- "scp-kde": Sequential split conformal prediction with Kernel Density Estimation of calibrated residuals
- "scp-bootstrap": Sequential split conformal prediction with independent bootstrap of calibrated residuals
- "scp-block-bootstrap": Sequential split conformal prediction with basic block bootstrap of calibrated residuals
- "scp2-kde": Sequential split conformal prediction with Kernel Density Estimation of standardized calibrated residuals
- "scp2-bootstrap": Sequential split conformal prediction with independent bootstrap of standardized calibrated residuals
- "scp2-block-bootstrap": Sequential split conformal prediction with basic block bootstrap of standardized calibrated residuals
block_size: int.
size of block for 'type_pi' in ("block-bootstrap", "scp-block-bootstrap", "scp2-block-bootstrap").
Default is round(3.15*(n_residuals^1/3))
replications: int.
number of replications (if needed, for predictive simulation). Default is 'None'.
Expand Down Expand Up @@ -186,6 +203,7 @@ def __init__(
type_scaling=("std", "std", "std"),
lags=1,
type_pi="kde",
block_size=None,
replications=None,
kernel=None,
agg="mean",
Expand Down Expand Up @@ -232,6 +250,7 @@ def __init__(
type_scaling=type_scaling,
seed=seed,
type_pi=type_pi,
block_size=block_size,
replications=replications,
kernel=kernel,
agg=agg,
Expand Down
48 changes: 34 additions & 14 deletions nnetsauce/lazypredict/lazyMTS.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -162,6 +162,7 @@ def __init__(
type_scaling=("std", "std", "std"),
lags=1,
type_pi="kde",
block_size=None,
replications=None,
kernel=None,
agg="mean",
Expand Down Expand Up @@ -194,6 +195,7 @@ def __init__(
backend=backend,
lags=lags,
type_pi=type_pi,
block_size=block_size,
replications=replications,
kernel=kernel,
agg=agg,
Expand All @@ -205,11 +207,11 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
Parameters:
X_train: array-like,
X_train: array-like or data frame,
Training vectors, where rows is the number of samples
and columns is the number of features.
X_test: array-like,
X_test: array-like or data frame,
Testing vectors, where rows is the number of samples
and columns is the number of features.
Expand Down Expand Up @@ -316,6 +318,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand All @@ -326,7 +329,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
),
]
)
else: # "random_state" in model().get_params().keys()
else: # "random_state" in model().get_params().keys()
pipe = Pipeline(
steps=[
("preprocessor", preprocessor),
Expand All @@ -347,6 +350,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand All @@ -365,9 +369,11 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):

X_pred = pipe["regressor"].predict(
h=X_test.shape[0], **kwargs
)
)

if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (
self.type_pi == "gaussian"
):
rmse = mean_squared_error(
X_test, X_pred.mean, squared=False
)
Expand All @@ -385,7 +391,9 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
MAE.append(mae)
MPL.append(mpl)

if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (
self.type_pi == "gaussian"
):
WINKLERSCORE.append(winklerscore)
COVERAGE.append(coveragecalc)
TIME.append(time.time() - start)
Expand All @@ -395,7 +403,9 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
CUSTOM_METRIC.append(custom_metric)

if self.verbose > 0:
if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (
self.type_pi == "gaussian"
):
scores_verbose = {
"Model": name,
"RMSE": rmse,
Expand Down Expand Up @@ -447,6 +457,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand All @@ -470,17 +481,18 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
seed=self.seed,
backend=self.backend,
show_progress=self.show_progress,
)

pipe.fit(X_train, xreg, **kwargs)
# pipe.fit(X_train, xreg=xreg) # DO xreg like in `ahead`

self.models[name] = pipe

if self.preprocess is True:
Expand All @@ -495,7 +507,9 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
h=X_test.shape[0], **kwargs
) # X_pred = pipe.predict(h=X_test.shape[0], new_xreg=new_xreg) ## DO xreg like in `ahead`

if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (
self.type_pi == "gaussian"
):
rmse = mean_squared_error(
X_test, X_pred.mean, squared=False
)
Expand All @@ -504,15 +518,19 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
winklerscore = winkler_score(X_pred, X_test, level=95)
coveragecalc = coverage(X_pred, X_test, level=95)
else:
rmse = mean_squared_error(X_test, X_pred.mean, squared=False)
rmse = mean_squared_error(
X_test, X_pred.mean, squared=False
)
mae = mean_absolute_error(X_test, X_pred.mean)
mpl = mean_pinball_loss(X_test, X_pred.mean)

names.append(name)
RMSE.append(rmse)
MAE.append(mae)
MPL.append(mpl)
if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (
self.type_pi == "gaussian"
):
WINKLERSCORE.append(winklerscore)
COVERAGE.append(coveragecalc)
TIME.append(time.time() - start)
Expand All @@ -522,7 +540,9 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
CUSTOM_METRIC.append(custom_metric)

if self.verbose > 0:
if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (
self.type_pi == "gaussian"
):
scores_verbose = {
"Model": name,
"RMSE": rmse,
Expand Down Expand Up @@ -554,7 +574,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
print(name + " model failed to execute")
print(exception)

if (self.replications is not None) or (self.type_pi=="gaussian"):
if (self.replications is not None) or (self.type_pi == "gaussian"):
scores = {
"Model": names,
"RMSE": RMSE,
Expand Down
12 changes: 10 additions & 2 deletions nnetsauce/lazypredict/lazydeepMTS.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -166,6 +166,7 @@ def __init__(
type_scaling=("std", "std", "std"),
lags=1,
type_pi="kde",
block_size=None,
replications=None,
kernel=None,
agg="mean",
Expand Down Expand Up @@ -199,6 +200,7 @@ def __init__(
backend=backend,
lags=lags,
type_pi=type_pi,
block_size=block_size,
replications=replications,
kernel=kernel,
agg=agg,
Expand All @@ -210,11 +212,11 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
Parameters:
X_train : array-like,
X_train : array-like or data frame,
Training vectors, where rows is the number of samples
and columns is the number of features.
X_test : array-like,
X_test : array-like or data frame,
Testing vectors, where rows is the number of samples
and columns is the number of features.
Expand Down Expand Up @@ -321,6 +323,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand Down Expand Up @@ -353,6 +356,7 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand Down Expand Up @@ -458,6 +462,8 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_clust=self.type_clust,
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand All @@ -480,6 +486,8 @@ def fit(self, X_train, X_test, xreg=None, **kwargs):
type_clust=self.type_clust,
type_scaling=self.type_scaling,
lags=self.lags,
type_pi=self.type_pi,
block_size=self.block_size,
replications=self.replications,
kernel=self.kernel,
agg=self.agg,
Expand Down
Loading

0 comments on commit 0f2984d

Please sign in to comment.