How to visualize decision tree in Python

Decision tree classifier is the most popularly used supervised learning algorithm. Unlike other classification algorithms, decision tree classifier in not a black box in the modeling phase.  What that’s means, we can visualize the trained decision tree to understand how the decision tree gonna work for the give input features.

So in this article, you are going to learn how to visualize the trained decision tree model in Python with graphviz. So let’s begin with the table of contents.

Table of contents

• Basic introduction to decision tree classifier
• Fruit classification with decision tree classifier
• Why we need to visualize the trained decision tree
• Understand the visualized decision tree
• Visualize decision tree in python
  • What is Graphviz
  • Visualize the decision tree online
  • Visualize the decision tree as pdf

Introduction to Decision tree classifier

Decision tree classifier is mostly used classification algorithm because of its advantages over other classification algorithms. When we say the advantages it’s not about the accuracy of the trained decision tree model. It’s all about the usage and understanding of the algorithm.

Decision tree advantages:

• Implementation wise building decision tree algorithm is so simple.
• The trained decision tree can use for both classification and regression problems.
• Complexity-wise decision tree is logarithmic in the number observation in the training dataset.
• The trained decision tree can visualize.

As we known the advantages of using the decision tree over other classification algorithms. Now let’s look at the basic introduction to the decision tree.

If you go through the article about the working of decision tree classifier in machine learning. You could aware of the decision tree keywords like root node, leaf node, information gain, gini index, tree pruning ..etc

The above keywords used to give you the basic introduction to decision tree classifier. If new to decision tree classifier, Please spend some time on the below articles before you continue reading about how to visualize the decision tree in Python.

• How the decision tree classifier works in machine learning

• Implementing decision tree classifier in Python with Scikit-Learn

• Building decision tree classifier in R programming language

Decision tree classifier is a classification model which creates set of rules from the training dataset. Later the created rules used to predict the target class. To get a clear picture of the rules and the need of visualizing decision, Let build a toy kind of decision tree classifier. Later use the build decision tree to understand the need to visualize the trained decision tree.

Fruit classification with decision tree classifier

The decision tree classifier will train using the apple and orange features, later the trained classifier can be used to predict the fruit label given the fruit features.

The fruit features is a dummy dataset. Below are the dataset features and the targets.

Weight (grams)	Smooth (Range of 1 to 10)	Fruit
170	9	1
175	10	1
180	8	1
178	8	1
182	7	1
130	3	0
120	4	0
130	2	0
138	5	0
145	6	0

 The dummy dataset having two features and target.

• Weight: Is the weight of the fruit in grams
• Smooth: Is the smoothness of the fruit in the range of 1 to 10
• Fruit: Is the target 1 means for apple and 0 means for orange.

Let’s follow the below workflow for modeling the fruit classifier.

• Loading the required Python machine learning packages
• Create and load the data in Pandas dataframe
• Building the fruit classifier with decision tree algorithm
• Predicting the fruit type from the trained classifier

Loading the required Python machine learning packages

 

 

The required python machine learning packages for building the fruit classifier are Pandas, Numpy, and Scikit-learn

• Pandas: For loading the dataset into dataframe, Later the loaded dataframe passed an input parameter for modeling the classifier.
• Numpy: For creating the dataset and for performing the numerical calculation.
• Sklearn: For training the decision tree classifier on the loaded dataset.

Now let’s create the dummy data set and load into the pandas dataframe

Create and load the data in Pandas dataframe

 

 

• The empty pandas dataframe created for creating the fruit data set.
• Using the numpy created arrays for target, weight, smooth.
  • The target having two unique values 1 for apple and 0 for orange.
  • Weight is the weight of the fruit in grams.
  • Smooth is the smoothness of the fruit in the range of 1 to 10.

Now, let’s use the loaded dummy dataset to train a decision tree classifier.

Building the fruit classifier with decision tree algorithm

 

 

• Creating the decision tree classifier instance from the imported scikit learn tree class.
• Using the loaded fruit data set features and the target to train the decision tree model.
• Print the trained fruit classifier.

Script Output:

 

Now let’s use the fruit classifier to predict the fruit type by giving the fruit features.

Predicting the fruit type from the trained classifier

 

 

Created 3 test data sets and using the trained fruit classifier to predict the fruit type and comparing with the real fruit type.

Script Output: 

 

The trained fruit classifier using the decision tree algorithm is accurately predicting the target fruit type for the given fruit features. You only know that decision tree is predicting the target fruit type for the given fruit features in a black box way and you don’t know what’s happing inside the black box.

To understand what happing inside the trained decision tree model and how it’s predicting the target class for the given features we need a visual representation of the trained decision tree classifier.

Why we need to visualize the trained decision tree

To answer the question why we need to visualize the trained decision tree, I am going to show you the visual representation of the above fruit classifier.

Before I show you the visual representation of the trained decision tree classifier, have a look at the 3 test observation we considered for predicting the target fruit type from the fruit classifier.

Weight	Smoot	Fruit	Classifer Predicted
170	9	1	1
180	10	1	1
130	2	0	0

The below image is the visual representation of the trained fruit classifier.

In the next coming section, you are going to learn how to visualize the decision tree in Python with graphviz.

Decision tree visualization explanation

The trained decision tree having the root node as fruit weight (x[0]). These conditions are populated with the provided train dataset.

If the weight is less than are equal to 157.5 go to the left node. If the weight is greater than 157.5 go to the right node. In fact, the right and left nodes are the leaf nodes as the decision tree considered only one feature (weight) is enough for classifying the fruit type.

The below pseudo code can represent the above graph into simple if else conditions.

 

 

Now if you pass the same 3 test observations we used to predict the fruit type from the trained fruit classifier you get to know why and how the trained decision tree predicting the fruit type for the given fruit features.

Visualize decision tree in python with graphviz

I hope you the advantages of visualizing the decision tree. Now let’s move the key section of this article, Which is visualizing the decision tree in python with graphviz.

You can visualize the trained decision tree in python with the help of graphviz. Below are two way to visualize the decision tree model.

• Visualize the decision tree online
• Visualize the decision tree as pdf

In both these case, you need first convert the trained decision tree classifier into graphviz object. Later we use the converted graphviz object for visualization. So It’s better to know about the python graphviz before looking into the visualization part.

What is Graphviz

Graphviz is one of the visualization libray. The gratness of graphviz is that it’s a open source visualiztion library. Graphiz widely used in networking applicaiton where to visulaze the connection beteen the swiths hub and differnt networks. When it’s comes to machine leanring used for decision tree and newral networks.

Now let’s look at how to visaulize the decision tree with graphviz.

Visualize the decision tree online

To visualize the decision tree online first you need to convert the trained decision tree, in our case the fruit classifier into a file (txt is better). Later you can use the contents of the converted file to visualize online.

The below can will convert the trained fruit classifier into graphviz object and saves into the txt file.

If you are having the proper python machine learning packages setup in your system. After running the above code fruit_classifier.txt will be saved on your local system.

To visualize the decision tree, you just need to open the fruit_classifier.txt file and copy the contents of the file to paste in the graphviz web portal. Below is the address for the web portal.

graphviz web portal address: http://webgraphviz.com

You can see the below graphviz web portal.

Once the graphviz web portal opened. Remove the already presented text in the text box and paste the text in the created txt file and click on the generate graph button.

For the modeled fruit classifier, we will get the below decision tree visualization.

Now let’s look at how to visualize the trained decision tree as pdf

Visualize the decision tree as pdf

The visualization of the trained decision tree as pdf will be same as the above. the only change is instead on copy and paste the contents of the converted txt file to the web portal, you will be converting it into a pdf file.

 

The above code will convert the trained decision tree classifier into graphviz object and then store the contents into the fruit_classifier.dot file. Next to convert the dot file into pdf file you can use the below command.

To preview the created pdf file you can use the below command.

You can get the complete code of this article on our Github account.