displaying scikit decision tree figure in jupyter notebook

Question

I am currently creating a machine learning jupyter notebook as a small project and wanted to display my decision trees. However, all options I can find are to export the graphics and then load a picture, which is rather complicated.

Therefore, I wanted to ask whether there is a way to display my decision trees directly without exporting and loading graphics.

Answer 1

As of scikit-learn version 21.0 (roughly May 2019), Decision Trees can now be plotted with matplotlib using scikit-learn’s tree.plot_tree without relying on graphviz.

import matplotlib.pyplot as plt
from sklearn.datasets import load_iris
from sklearn.tree import DecisionTreeClassifier
from sklearn import tree

X, y = load_iris(return_X_y=True)

# Make an instance of the Model
clf = DecisionTreeClassifier(max_depth = 5)

# Train the model on the data
clf.fit(X, y)

fn=['sepal length (cm)','sepal width (cm)','petal length (cm)','petal width (cm)']
cn=['setosa', 'versicolor', 'virginica']

# Setting dpi = 300 to make image clearer than default
fig, axes = plt.subplots(nrows = 1,ncols = 1,figsize = (4,4), dpi=300)

tree.plot_tree(clf,
           feature_names = fn, 
           class_names=cn,
           filled = True);

# You can save your plot if you want
#fig.savefig('imagename.png')

Something similar to what is below will output in your jupyter notebook.

The code was adapted from this post.

Answer 2

You can show the tree directly using IPython.display:

import graphviz
from sklearn.tree import DecisionTreeRegressor, DecisionTreeClassifier,export_graphviz
from sklearn.datasets import make_regression

# Generate a simple dataset
X, y = make_regression(n_features=2, n_informative=2, random_state=0)
clf = DecisionTreeRegressor(random_state=0, max_depth=2)
clf.fit(X, y)
# Visualize the tree
from IPython.display import display
display(graphviz.Source(export_graphviz(clf)))