rfflearn.cpu.permutation_feature_importance

Calculate permutation importance, and set the feature importance (mean of permutation importance for each trial) as model.feature_importances_.

def rfflearn.cpu.permutation_feature_importance(model, Xs, ys, **kwargs):

Parameters:

model: object

An estimator that has already been fitted.

Xs: np.ndarray

Input data of shape (n_samples, n_features), on which permutation importance will be computed.

ys: np.ndarray

Input label of shape (n_samples,) or (n_samples, n_classes). Specify None for unsupervised model.

kwargs: dict  (default={})

Additional keyword arguments for sklearn.inspection.permutation_importance.

Returns:

np.ndarray

Raw permutation importance scores of shape (n_features, n_repeats).

Notes:

This function add an attribute feature_importances_ to the model instance.

rfflearn.cpu.permutation_plot

Visualize permutation importance as a box diagram.

def rfflearn.cpu.permutation_plot(permutation_importances, feature_names, show=True):

Parameters:

permutation_importances: np.ndarray

Raw scores of shape (num_features, num_repeats).

feature_names: list

List of feature names.

show: bool

Shows plot if True.

rfflearn.cpu.shap_feature_importance

def rfflearn.cpu.shap_feature_importance(model, Xs):

Parameters:

model: object

An estimator that has already been fitted.

Xs: np.ndarray

Input data of shape (n_samples, n_features), on which permutation importance will be computed.

Returns:

np.ndarray

SHAP values of shape (n_samples, n_features).

Notes:

This function add an attribute feature_importances_ to the model instance.

rfflearn.cpu.shap_plot

Calculate SHAP values using shap library, and set the feature importance (absolute of SHAP values) as model.feature_importances_.

def rfflearn.cpu.shap_plot(*pargs, **kwargs):

Parameters:

pargs: tuple  (default: ())

Positional arguments for shap.summary_plot.

kwargs: dict  (default: {})

Keyword arguments for shap.summary_plot.

Notes:

This function is just an alias of the shap.summary_plot function because the shap.summary_plot is implemented very well!!

Minimal Example

>>> import numpy as np                                        # Import Numpy
>>> import rfflearn.cpu as rfflearn                           # Import rfflearn
>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])        # Define input data
>>> y = np.array([1, 1, 2, 2])                                # Definr label data
>>> model = rfflearn.RFFSVC().fit(X, y)                       # Training a model
>>> Is = rfflearn.permutation_feature_importance(model, X, y) # Compute permutation importance
>>> model.feature_importances_                                # Confirm feature importance
array([0.25, 0.3 ])
>>> permutation_plot(Is, ["x1", "x2"], show=True)             # Plot importance and show