gitea_admin/suse-model-store
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Release 11 (from main)

Browse Source
This commit is contained in:
model-plugin-warehouse CI
2026-10-23 18:04:02 +00:00
parent dc12d4e35c
commit efe522474f
11 changed files with 103 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
models/linear_regression/README.md
View File

@@ -23,7 +23,7 @@ These parameters are defined in the `model.yaml` under `data_model.input_map` an
| `scaler__name` | `'Standard Scaler'` | Name of the scaler to use for scaling the data. Options: 'Standard Scaler' |
| `verbose` | `true` | Enable verbose output during model execution. |
| `steps_order` | `['Discontinuity Treatment', 'Lag Selection', 'Range Selection & Data Removal', 'Static Window Removal', 'Define Variables Limits', 'Normalization', 'Feature Creation', 'Lag Creation']` | Order of the steps to be executed in the pipeline. |
| `nan_treatment` | `'drop'` | The treatment for missing values. Options: 'drop', 'linear interpolation' |
| `nan_treatment` | `'drop'` | The treatment for missing values. Options: 'drop' (default), 'linear interpolation', 'fill linear' (alias) |
| `lag_train` | `{'variable_name': 1}` | The lags for each variable to be applyed during training. |
| `lag_transform` | `{'variable_name': 1}` | The lags for each variable to be applyed during transformation. |
| `start_date` | `'2020-01-01 00:00:00'` | The start date for the dataset. Format: 'YYYY-MM-DD HH:MM:SS' |
@@ -58,6 +58,7 @@ Neither method mutates the passed DataFrames.
- If `variable_columns` is not set, it is inferred during `fit` as all columns except the target.
- Polynomial features are created automatically when `degree > 1`.
- Rows with NaN in any model input get a NaN prediction (the output always has one value per input row); rows with NaN features or NaN target are left out of the fit.
- `clipping` uses Q1/Q3 of the training target when replacing out-of-bounds predictions.
- The wrapper enforces that `wrapper.target` is set before training or retraining; otherwise, a `ValueError` is raised to fail fast when the configuration is incomplete.

4
models/linear_regression/index.yaml
View File

@@ -8,8 +8,8 @@ requirements:
- cloudpickle==3.1.2
- mlflow==3.8.1
- git+ssh://git@github.com/Aignosi/sientia-dataops-library.git@1.11.0
- git+ssh://git@github.com/Aignosi/sientia-model-library.git@0.14.0
version: 1.2.0
- git+ssh://git@github.com/Aignosi/sientia-model-library.git@0.14.3
version: 1.2.2
type: model
model:
class: LinearRegressionModel

126
models/linear_regression/linear_regression_model.py
View File

@@ -64,6 +64,15 @@ class LinearRegressionModel(BaseEstimator, TransformerMixin):
"""
return self.regr
def _check_model_features(self, columns: pd.Index, model_features: list) -> None:
"""Raise when columns lack any feature the regressor was fitted with."""
missing_model_features = [c for c in model_features if c not in columns]
if missing_model_features:
raise ValueError(
'Prediction data is missing required transformed feature columns: '
+ ', '.join(missing_model_features)
)
def create_poly_features(self, input_data: pd.DataFrame, fit: bool = False) -> pd.DataFrame:
"""
Function to create polynomial features
@@ -110,15 +119,55 @@ class LinearRegressionModel(BaseEstimator, TransformerMixin):
return input_data
def _fittable_rows(
self, features: pd.DataFrame, target: pd.Series
) -> tuple[pd.DataFrame, pd.Series]:
"""
Keep only what LinearRegression can fit: inf becomes NaN, all-null columns are dropped
(degree 1 only) and rows with a NaN feature or a NaN target are removed.
Raises:
ValueError: If no rows are left.
"""
# Handle infinite values
features = features.astype(float).replace([np.inf, -np.inf], np.nan)
# Drop columns with all null values. With degree > 1 they are kept (PolynomialFeatures
# needs every variable column), so the row mask below leaves no rows and fit raises.
if self.degree == 1:
all_null_cols = features.columns[features.isnull().all()].tolist()
features = features.drop(columns=all_null_cols)
if self.verbose and all_null_cols:
print(f'{len(all_null_cols)} columns with all null values were dropped.')
# Rows with NaN in any feature or in the target cannot be fitted (SIENTIAPDE-2215).
# Masked before PolynomialFeatures, which rejects NaN.
mask = features.notna().all(axis=1) & target.notna()
features = features[mask]
target = target[mask]
if features.empty:
raise ValueError(
'No training rows left after removing rows with NaN features or NaN target.'
)
if self.verbose:
print(f'X_train shape after removing NaN rows: {features.shape}')
print(f'y_train shape after removing NaN rows: {target.shape}')
return features, target
def fit(self, input_data: pd.DataFrame) -> 'LinearRegressionModel':
"""
Function to fit the model
Function to fit the model. Rows with NaN features or a NaN target are left out of the fit.
Args:
input_data (pandas.DataFrame): The data used to fit the Linear Regression model
Returns:
LinearRegressionModel: The prediction model object
Raises:
ValueError: If target_variable is unset, a feature column is missing, or no
training rows are left after removing rows with NaN features or NaN target.
KeyError: If the target column is missing from input_data.
"""
if self.verbose:
# Display the header
@@ -153,30 +202,14 @@ class LinearRegressionModel(BaseEstimator, TransformerMixin):
print(f'X_train shape: {X_train.shape}')
print(f'y_train shape: {y_train.shape}')
X_train, y_train = self._fittable_rows(X_train, y_train)
# Create polynomial features
if self.degree > 1:
X_train = self.create_poly_features(X_train, fit=True)
# Handle infinite values
X_train = X_train.replace([np.inf, -np.inf], np.nan)
# Drop columns with all null values
all_null_cols = X_train.columns[X_train.isnull().all()].tolist()
X_train = X_train.drop(columns=all_null_cols)
if self.verbose and all_null_cols:
print(f'{len(all_null_cols)} columns with all null values were dropped.')
# Get mask of rows with any NaN values
mask = X_train.notna().all(axis=1)
X_train = X_train[mask]
y_train = y_train[mask]
if self.verbose:
print(f'X_train shape after removing NaN rows: {X_train.shape}')
print(f'y_train shape after removing NaN rows: {y_train.shape}')
self.q1_target = y_train.quantile(0.25)
self.q3_target = y_train.quantile(0.75)
self.q1_target = float(y_train.quantile(0.25))
self.q3_target = float(y_train.quantile(0.75))
# Fit the model
self.regr.fit(X_train, y_train)
@@ -211,7 +244,7 @@ class LinearRegressionModel(BaseEstimator, TransformerMixin):
input_data (pandas.DataFrame): The data used to predict the target variable
Returns:
numpy.ndarray: The predicted target variable
numpy.ndarray: One prediction per input row; NaN where any model input is NaN
"""
if self.verbose:
# Display the header
@@ -228,38 +261,41 @@ class LinearRegressionModel(BaseEstimator, TransformerMixin):
+ ', '.join(missing_features)
)
X_test = input_data[self.variable_columns]
# astype(float) first: laborious sends object columns with None, and replace() on
# object dtype raises a pandas downcasting FutureWarning on every call.
X_test = input_data[self.variable_columns].astype(float).replace([np.inf, -np.inf], np.nan)
if self.verbose:
print(f'X_test shape: {X_test.shape}')
# Create polynomial features
model_features = list(self.regr.feature_names_in_)
# Rows with any NaN model input get a NaN prediction, so the output keeps the input
# length (SIENTIAPDE-2215). With degree > 1 the mask uses the raw variable columns,
# because PolynomialFeatures rejects NaN.
if self.degree > 1:
X_test = self.create_poly_features(X_test, fit=False)
X_test = X_test.replace([np.inf, -np.inf], np.nan)
row_mask = X_test.notna().all(axis=1)
else:
self._check_model_features(X_test.columns, model_features)
row_mask = X_test[model_features].notna().all(axis=1)
model_features = self.regr.feature_names_in_
# Ensure required model features exist
missing_model_features = [c for c in model_features if c not in X_test.columns]
if missing_model_features:
raise ValueError(
'Prediction data is missing required transformed feature columns: '
+ ', '.join(missing_model_features)
)
y_pred = np.full(len(input_data), np.nan)
if not row_mask.any():
return y_pred
# Get mask of rows with any NaN values
mask = X_test.notna().all(axis=1)
X_test = X_test[mask]
x_valid = X_test[row_mask]
if self.degree > 1:
x_valid = self.create_poly_features(x_valid, fit=False)
self._check_model_features(x_valid.columns, model_features)
if self.verbose:
print(f'X_test shape after removing NaN rows: {X_test.shape}')
print(f'X_test shape after removing NaN rows: {x_valid.shape}')
# For simplicity and to keep behavior consistent with prior tests, drop NaN rows before prediction
y_pred = self.regr.predict(X_test[model_features])
valid_pred = self.regr.predict(x_valid[model_features])
if self.clipping:
for i in range(len(y_pred)):
if y_pred[i] > self.clipping['max']:
y_pred[i] = self.q3_target
elif y_pred[i] < self.clipping['min']:
y_pred[i] = self.q1_target
for i in range(len(valid_pred)):
if valid_pred[i] > self.clipping['max']:
valid_pred[i] = self.q3_target
elif valid_pred[i] < self.clipping['min']:
valid_pred[i] = self.q1_target
y_pred[row_mask.to_numpy()] = valid_pred
return y_pred

4
models/linear_regression/schemas.yaml
View File

@@ -61,10 +61,10 @@ components:
description: "Order of the steps to be executed in the pipeline."
nan_treatment:
type: string
enum: ["drop", "linear interpolation"]
enum: ["drop", "linear interpolation", "fill linear"]
default: "drop"
example: "drop"
description: "The treatment for missing values."
description: "The treatment for missing values. 'fill linear' is an alias of 'linear interpolation'."
lag_train:
type: object
x-feature: true