Python - export_graphviz class_name type error

Question

I am actively learning how to implement decision trees in python.

When recreating the Iris classification example from scikit-learn, i get a TypeError for parameters that exist in export_graphviz, namely 'class_names' and 'plot_options'.

from IPython.display import Image  
import sklearn
dot_data = StringIO()  
sklearn.tree.export_graphviz(clf, out_file=dot_data,
                     plot_options=['class', 'filled', 'label', 'sample',       'proportion'],
                 target_names=iris['target_names'],
                 feature_names=iris['feature_names'])
graph = pydot.graph_from_dot_data(dot_data.getvalue())  
Image(graph.create_png())

The error for the specific code above is:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-33-aba117838252> in <module>()
      5                      plot_options=['class', 'filled', 'label', 'sample', 'proportion'],
      6                      target_names=iris['target_names'],
----> 7                      feature_names=iris['feature_names'])
      8 graph = pydot.graph_from_dot_data(dot_data.getvalue())
      9 Image(graph.create_png())

TypeError: export_graphviz() got an unexpected keyword argument 'plot_options'

On my computer, I have graphviz and pydot2 installed. I receive an error when trying to install pygraphviz:

 If you think your installation is correct you will need to manually

    change the include_dirs and library_dirs variables in setup.py to

    point to the correct locations of your graphviz installation.



    The current setting of library_dirs and include_dirs is:

library_dirs=None

include_dirs=None

error: Error locating graphviz.

Is there a work around/solution to allow me to use the parameters in export_graphviz in order to build the tree visualization I want? Would pursuing a solution to the pygraphviz install error lead to a solution for my tree?

Thank you,

Where did you see plot_options, target_names, etc in the following example? And what do you want to visualise exactly? — Ibraim Ganiev
– Ibraim Ganiev, Commented Nov 19, 2015 at 5:29

mujjiga · Accepted Answer · 2016-01-05 17:04:37Z

The signature of export_graphviz is

def export_graphviz(decision_tree, out_file="tree.dot", max_depth=None,
                feature_names=None, class_names=None, label='all',
                filled=False, leaves_parallel=False, impurity=True,
                node_ids=False, proportion=False, rotate=False,
                rounded=False, special_characters=False):

The correct function call is (assuming that your data is in iris object)

sklearn.tree.export_graphviz(clf, out_file=dot_data,  
                     feature_names=iris['feature_names'],  
                     class_names=iris['target_names'],  
                     filled=True, rounded=True,  
                     special_characters=True)

In case it throws up the error

TypeError: export_graphviz() got an unexpected keyword argument 'class_names'

it means your sklearn is of older version. If you are using anaconda python distribution you can update to latest version of sklearn using conda update scikit-learn. Use pip command if you are on any other distribution. Make sure your sklearn is 0.17 version.

import sklearn
print sklearn.__version__

Collectives™ on Stack Overflow

Python - export_graphviz class_name type error

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related