Module linear_model (2.25.0)

Linear models. This module is styled after scikit-learn's linear_model module: https://scikit-learn.org/stable/modules/linear_model.html.

Classes

LinearRegression

LinearRegression(
    *,
    optimize_strategy: typing.Literal[
        "auto_strategy", "batch_gradient_descent", "normal_equation"
    ] = "auto_strategy",
    fit_intercept: bool = True,
    l1_reg: typing.Optional[float] = None,
    l2_reg: float = 0.0,
    max_iterations: int = 20,
    warm_start: bool = False,
    learning_rate: typing.Optional[float] = None,
    learning_rate_strategy: typing.Literal["line_search", "constant"] = "line_search",
    tol: float = 0.01,
    ls_init_learning_rate: typing.Optional[float] = None,
    calculate_p_values: bool = False,
    enable_global_explain: bool = False
)

Ordinary least squares Linear Regression.

LinearRegression fits a linear model with coefficients w = (w1, ..., wp) to minimize the residual sum of squares between the observed targets in the dataset, and the targets predicted by the linear approximation.

Examples:

>>> from bigframes.ml.linear_model import LinearRegression
>>> import bigframes.pandas as bpd
>>> bpd.options.display.progress_bar = None
>>> X = bpd.DataFrame({                 "feature0": [20, 21, 19, 18],                 "feature1": [0, 1, 1, 0],                 "feature2": [0.2, 0.3, 0.4, 0.5]})
>>> y = bpd.DataFrame({"outcome": [0, 0, 1, 1]})
>>> # Create the linear model
>>> model = LinearRegression()
>>> model.fit(X, y)
LinearRegression()

>>> # Score the model
>>> score = model.score(X, y)
>>> print(score) # doctest:+SKIP
    mean_absolute_error  mean_squared_error  mean_squared_log_error          0             0.022812            0.000602                 0.00035
    median_absolute_error  r2_score  explained_variance
0               0.015077  0.997591            0.997591
Parameters
Name Description
optimize_strategy str, default "auto_strategy"

The strategy to train linear regression models. Possible values are "auto_strategy", "batch_gradient_descent", "normal_equation". Default to "auto_strategy".
fit_intercept bool, default True

Default True. Whether to calculate the intercept for this model. If set to False, no intercept will be used in calculations (i.e. data is expected to be centered).
l1_reg float or None, default None

The amount of L1 regularization applied. Default to None. Can't be set in "normal_equation" mode. If unset, value 0 is used.
l2_reg float, default 0.0

The amount of L2 regularization applied. Default to 0.
max_iterations int, default 20

The maximum number of training iterations or steps. Default to 20.
warm_start bool, default False

Determines whether to train a model with new training data, new model options, or both. Unless you explicitly override them, the initial options used to train the model are used for the warm start run. Default to False.
learning_rate float or None, default None

The learn rate for gradient descent when learning_rate_strategy='constant'. If unset, value 0.1 is used. If learning_rate_strategy='line_search', an error is returned.
learning_rate_strategy str, default "line_search"

The strategy for specifying the learning rate during training. Default to "line_search".
tol float, default 0.01

The minimum relative loss improvement that is necessary to continue training when EARLY_STOP is set to true. For example, a value of 0.01 specifies that each iteration must reduce the loss by 1% for training to continue. Default to 0.01.
ls_init_learning_rate float or None, default None

Sets the initial learning rate that learning_rate_strategy='line_search' uses. This option can only be used if line_search is specified. If unset, value 0.1 is used.
calculate_p_values bool, default False

Specifies whether to compute p-values and standard errors during training. Default to False.
enable_global_explain bool, default False

Whether to compute global explanations using explainable AI to evaluate global feature importance to the model. Default to False.

LogisticRegression

LogisticRegression(
    *,
    optimize_strategy: typing.Literal[
        "auto_strategy", "batch_gradient_descent"
    ] = "auto_strategy",
    fit_intercept: bool = True,
    l1_reg: typing.Optional[float] = None,
    l2_reg: float = 0.0,
    max_iterations: int = 20,
    warm_start: bool = False,
    learning_rate: typing.Optional[float] = None,
    learning_rate_strategy: typing.Literal["line_search", "constant"] = "line_search",
    tol: float = 0.01,
    ls_init_learning_rate: typing.Optional[float] = None,
    calculate_p_values: bool = False,
    enable_global_explain: bool = False,
    class_weight: typing.Optional[
        typing.Union[typing.Literal["balanced"], typing.Dict[str, float]]
    ] = None
)

Logistic Regression (aka logit, MaxEnt) classifier.

from bigframes.ml.linear_model import LogisticRegression import bigframes.pandas as bpd bpd.options.display.progress_bar = None X = bpd.DataFrame({ "feature0": [20, 21, 19, 18], "feature1": [0, 1, 1, 0], "feature2": [0.2, 0.3, 0.4, 0.5]}) y = bpd.DataFrame({"outcome": [0, 0, 1, 1]})

Create the LogisticRegression

model = LogisticRegression() model.fit(X, y) LogisticRegression() model.predict(X) # doctest:+SKIP predicted_outcome predicted_outcome_probs feature0 feature1 feature2 0 0 [{'label': 1, 'prob': 3.1895929877221615e-07} ... 20 0 0.2 1 0 [{'label': 1, 'prob': 5.662891265051953e-06} ... 21 1 0.3 2 1 [{'label': 1, 'prob': 0.9999917826885262} {'l... 19 1 0.4 3 1 [{'label': 1, 'prob': 0.9999999993659574} {'l... 18 0 0.5 4 rows × 5 columns

[4 rows x 5 columns in total]

Score the model

score = model.score(X, y) score # doctest:+SKIP precision recall accuracy f1_score log_loss roc_auc 0 1.0 1.0 1.0 1.0 0.000004 1.0 1 rows × 6 columns

[1 rows x 6 columns in total]

Parameters
Name Description
optimize_strategy str, default "auto_strategy"

The strategy to train logistic regression models. Possible values are "auto_strategy" and "batch_gradient_descent". The two are equilevant since "auto_strategy" will fall back to "batch_gradient_descent". The API is kept for consistency. Default to "auto_strategy".
fit_intercept default True

Default True. Specifies if a constant (a.k.a. bias or intercept) should be added to the decision function.
class_weight dict or 'balanced', default None

Default None. Weights associated with classes in the form {class_label: weight}.If not given, all classes are supposed to have weight one. The "balanced" mode uses the values of y to automatically adjust weights inversely proportional to class frequencies in the input data as n_samples / (n_classes * np.bincount(y)). Dict isn't supported.
l1_reg float or None, default None

The amount of L1 regularization applied. Default to None. Can't be set in "normal_equation" mode. If unset, value 0 is used.
l2_reg float, default 0.0

The amount of L2 regularization applied. Default to 0.
max_iterations int, default 20

The maximum number of training iterations or steps. Default to 20.
warm_start bool, default False

Determines whether to train a model with new training data, new model options, or both. Unless you explicitly override them, the initial options used to train the model are used for the warm start run. Default to False.
learning_rate float or None, default None

The learn rate for gradient descent when learning_rate_strategy='constant'. If unset, value 0.1 is used. If learning_rate_strategy='line_search', an error is returned.
learning_rate_strategy str, default "line_search"

The strategy for specifying the learning rate during training. Default to "line_search".
tol float, default 0.01

The minimum relative loss improvement that is necessary to continue training when EARLY_STOP is set to true. For example, a value of 0.01 specifies that each iteration must reduce the loss by 1% for training to continue. Default to 0.01.
ls_init_learning_rate float or None, default None

Sets the initial learning rate that learning_rate_strategy='line_search' uses. This option can only be used if line_search is specified. If unset, value 0.1 is used.
calculate_p_values bool, default False

Specifies whether to compute p-values and standard errors during training. Default to False.
enable_global_explain bool, default False

Whether to compute global explanations using explainable AI to evaluate global feature importance to the model. Default to False.