Introduction

I am studying using "100 Python Practice Knock". I think this book is really good because you can study data analysis using data that is close to the actual situation (although there are actually a lot of even worse data ...). In Chapter 5 of this book, there is a scene where scikit-learn is used for analysis using a decision tree. So I created a model, but I tried to visualize the tree structure using graphviz, so I tried it this time.

Because beginners have posted, I would appreciate it if you could teach me if you make a mistake.

Target readers

--Those who are reading 100 Python practice knocks --Those who are looking for a visualization method of decision trees with scikit-learn

Target location

Python practice 100 knocks -> Chapter 5 10 knocks to predict customer withdrawal -> Knock 49: Let's check the variables that the model contributes

My pc environment

macOS Catalina 10.15.3
Python 3.7.7
scikit-learn 0.22.2.post1
Jupyter 1.0.0

installation of graphviz

First, install the main body using homebrew.

brew install graphviz

In addition, install the library for Python using pip (anaconda seems to be able to do it with conda).

pip install graphviz

code

The following is based on the assumption that the code on p.127 has been implemented.

1. How to display on Jupyter notebook

All you have to do is add the following code. It's very easy.

from sklearn import tree
import graphviz
dot_data = tree.export_graphviz(model, out_file=None)
graph = graphviz.Source(dot_data)
graph

2. How to create a file

You can create a pdf file with just a few modifications to the last code. In this example, "test.pdf" is created in the current directory.

from sklearn import tree
import graphviz
dot_data = tree.export_graphviz(model, out_file=None)
graph = graphviz.Source(dot_data)
graph.render('test')

3. How to create a file using the terminal (bonus)

I used the export_graphviz function of sklearn.tree to create a decision tree file in DOT language format and run system commands on Jupyter notebook.

from sklearn import tree
import graphviz
with open('test.dot', mode='w') as f:
    tree.export_graphviz(model, out_file=f)
!dot -T png test.dot -o test.png

References

This article was written with reference to the following information. See below for more information. 1.10. Decision Trees (official document) sklearn.tree.export_graphviz (also official documentation) Try the Decision Tree with Python: scikit-learn (I used it as a reference)

Python practice 100 knocks I tried to visualize the decision tree of Chapter 5 using graphviz