Implementation:DistrictDataLabs Yellowbrick FeatureImportances Visualizer

Knowledge Sources	Yellowbrick Yellowbrick Docs
Domains	Machine_Learning, Model_Selection, Visualization
Last Updated	2026-02-08 00:00 GMT

Overview

Concrete tool for visualizing feature importance rankings as horizontal bar charts, provided by the Yellowbrick library.

Description

The FeatureImportances visualizer displays a horizontal bar chart of features ranked by their importance as determined by a fitted scikit-learn estimator. It works with any model that exposes either a feature_importances_ attribute (tree-based models) or a coef_ attribute (linear models) after fitting. The visualizer searches for these attributes in order, using whichever is found first.

The class extends ModelVisualizer from the Yellowbrick base module. When fit(X, y) is called, the visualizer first fits the underlying estimator (unless it is already fitted, controlled by the is_fitted parameter), then extracts the feature importances. For multi-class classifiers with multi-dimensional coef_ arrays of shape (n_classes, n_features), the visualizer either computes the mean across classes (when stack=False) or produces a stacked bar chart showing per-class contributions (when stack=True). Importances can be displayed as absolute values via the absolute parameter, and normalized relative to the strongest feature via the relative parameter. The topn parameter allows displaying only the top N (positive integer) or bottom N (negative integer) features.

Feature labels are automatically extracted from DataFrame column names when a DataFrame is passed, or can be explicitly provided via the labels parameter.

Usage

Use this visualizer for feature engineering and model interpretation. It is appropriate when you need to identify which features are most influential, perform feature selection, or communicate model behavior to non-technical stakeholders.

Code Reference

Source Location

• Repository: yellowbrick
• File: yellowbrick/model_selection/importances.py
• Class Lines: L38-406 (class), L124-138 (__init__), L154-267 (fit)
• Quick Method Lines: L413-531

Signature

class FeatureImportances(ModelVisualizer):
    def __init__(
        self,
        estimator,
        ax=None,
        labels=None,
        relative=True,
        absolute=False,
        xlabel=None,
        stack=False,
        colors=None,
        colormap=None,
        is_fitted="auto",
        topn=None,
        **kwargs
    ):

Import

from yellowbrick.model_selection import FeatureImportances

I/O Contract

Inputs

Name	Type	Required	Description
estimator	scikit-learn estimator	Yes	A model that exposes coef_ or feature_importances_ after fitting.
ax	matplotlib.Axes	No	The axes object to plot on. Default: None (current axes).
labels	list	No	Feature names. If None, uses DataFrame column names or integer indices.
relative	bool	No	Show features as percentage of strongest. Default: True.
absolute	bool	No	Take absolute value of coefficients. Default: False.
xlabel	str	No	Custom label for x-axis. Default: None (auto-determined).
stack	bool	No	If True and multi-class, show stacked bar chart. Default: False.
colors	list of strings	No	Colors for each bar when stack=False.
colormap	string or matplotlib cmap	No	Colormap for classes when stack=True.
is_fitted	bool or str	No	Whether the estimator is already fitted. Default: "auto".
topn	int	No	Show only top N (positive) or bottom N (negative) features. Default: None (all).

The fit(X, y) method accepts:

Name	Type	Required	Description
X	ndarray or DataFrame, shape (n, m)	Yes	Feature matrix of n instances with m features.
y	ndarray or Series, length n	No	Target or class values.

Outputs

Name	Type	Description
features_	np.array	Feature labels ranked according to their importance.
feature_importances_	np.array	Numeric importance values computed by the model (optionally relative and/or absolute).
classes_	np.array	Class labels for classifiers. None for regressors.

Usage Examples

Basic Usage

from sklearn.ensemble import GradientBoostingClassifier
from yellowbrick.model_selection import FeatureImportances

# Create and fit the visualizer
viz = FeatureImportances(GradientBoostingClassifier(), relative=True, topn=10)
viz.fit(X_train, y_train)
viz.show()

Quick Method

from yellowbrick.model_selection import feature_importances
from sklearn.ensemble import RandomForestClassifier

feature_importances(RandomForestClassifier(), X_train, y_train, topn=5)

Related Pages

Implements Principle

• Principle:DistrictDataLabs_Yellowbrick_Feature_Importance_Ranking

Requires Environment

• Environment:DistrictDataLabs_Yellowbrick_Python_Scikit_Learn_Environment