Decision tree classification is a popular supervised machine learning algorithm and frequently used to classify categorical data as well as regressing continuous data. In this article, we will learn how can we implement decision tree classification using Scikit-learn package of Python
Decision tree classification helps to take vital decisions in banking and finance sectors like whether a credit/loan should be given to a customer or not depending on his risk bearing credentials; in medical test conditions like if a new medicine should be tried on a patient depending on his/her medical history and many more fields.
- Machine learning with python: a detailed discussion
- Supervised machine learning
- Unsupervised machine learning
The above two cases are where the target variable is a bivariate one i.e. with only two categories of response. There can be cases where the target variable has more than two categories, the decision tree can be applied in such multinomial cases too. The decision tree can also handle both numerical and categorical data. So, no doubt a decision tree gives a lot of liberty to its users.
NB: Being a non-native English speaker, I always take extra care to proofread my articles with Grammarly. It is the best grammar and spellchecker available online. Read here my review of using Grammarly for more than two years.
Introduction to decision tree
Page contents
Decision tree problems generally consist of some existing conditions which determine its categorical response. If we arrange the conditions and the decisions depending on those conditions and again one of those decisions resulting in further decisions; the whole structure of decision making resembles a tree structure. Hence the name decision tree.
The first and topmost condition which initiates the decision-making process is called the root condition. The nodes from the root node are called either a leaf node or decision node depending on which one takes part in further decision making. In this way, a recursive process of continues unless and until all the elements are grouped into particular categories and final nodes are all leaf nodes.
An example of decision tree
Here we can take an example of recent COVID-19 epidemic problem related to the testing of positive cases. We all know that the main problem with this disease is that it is very infectious. So, to identify COVID positive patients and isolating them is very essential to stop its further spread. This needs rigorous testing. But COVID testing is a time consuming and resource-intensive process. It becomes more of a challenge in the case of countries like India with a strong 1.3 billion population.
So, if we can categorize which persons actually need testing it can save a lot of time and resources. We can straightway downsize the testing population significantly. So, it is a kind of divide and conquer policy. See the below decision tree for classifying persons who need to be tested.
The whole classification process is much similar to how a human being judges a situation and makes a decision. That’s why this machine learning technique is simple to understand and easier to implement. Further being a non-parametric approach this algorithm is applicable to any kind of data even when the distribution is not known.
The distinct character of a decision tree which makes it special among all other machine learning algorithms is that unlike them it is a white box technique. That means the logic used in the classification process is visible to us. Due to simple logic, the training time for this algorithm is far less even when the data size is huge with high dimensionality. Moreover, it is the decision tree which makes the foundation of advanced machine learning computing technique like the random forest, bagging, gradient boosting etc.
Advantages of decision tree
- The decision tree has a great advantage of being capable of handling both numerical and categorical variables. Many other modelling techniques can handle only one kind of variable.
- No data preprocessing is required. Except for missing values no other data processing steps like data standardization, use of dummy variables for categorical data are required for decision tree which saves a lot of user’s time.
- The assumptions are not too rigid and model can slightly deviate from them.
- The decision tree model validation can be done through statistical tests and the reliability can be established easily.
- As it is a white box model, so the logic behind it is visible to us and we can easily interpret the result unlike the black-box model like an artificial neural network.
Now no technique can be without any flaws, there are always some flipside and decision tree is no exception.
Disadvantages of Decision tree
- A very serious problem with a decision tree is that it is very much prone to overfitting. That means the prediction given by decision tree is often too accurate for a too specific situation with a too complex model.
- The classification by decision tree generally uses an algorithm which tends to find a local optimum result for each node. As this process follows recursively for each node, ultimately the whole process ends up finding a locally optimal instead of a globally optimal decision tree.
- The result obtained from a decision tree is very unstable. A little variation in the data can lead to a completely different classification/regression result. That’s why the concept of random forest/ensemble technique came, this technique brings together the best result obtained from a number of models instead of relying on a single one.
Classification and Regression Tree (CART)
The decision tree has two main categories classification tree and regression tree. These two terms at a time called as CART. This term was first coined in 1984 by Leo Breiman, Jerome Friedman, Richard Olshen and Charles Stone.
Classification
When the response is categorical in nature, the decision tree performs classification. Like the examples, I gave before, whether a person is sick or not or a product is pass or fail in a quality test. In all these cases the problem in hand is to include the target variable into a group.
The target variable can be a binomial that is with only two categories like yes-no, male-female, sick-not sick etc. or the target variable can be multinomial that is with more than two categories. An example of a multinomial variable can be the economic status of people. It can have categories like very rich, rich, middle class, lower-middle class, poor, very poor etc. Now the benefit of the decision tree is a decision tree is capable of handling both binomial and multinomial variables.
Regression
On the other hand, the decision tree has its application in regression problem when the target variable is of continuous nature. For example, predicting the rainfall of a future date depending on other weather parameters. Here the target variable is a continuous one. So, it is a problem of regression.
Application of Decision tree with Python
Here we will use the sci-kit learn package to implement the decision tree. The package has a function called DecisionTreeClasifier() which is capable of classifying both binomial (target variable with only two classes) and multinomial (target variable having more than two classes) variables.
Performing classification using decision tree
Importing required libraries
The first step to start coding is to import all the libraries we are going to use. The basic libraries for any kind of data science projects are like pandas, numpy, matplotlib etc. The purpose of these libraries has an elaborate discussion in the article simple linear regression with python.
# importing libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
About the data
The example dataset I have used here for demonstration purpose is from kaggle.com. The data collected by “National Institute of Diabetes and Digestive and Kidney Diseases” contains vital parameters of diabetes patients belong to Pima Indian heritage.
Here is a glimpse of the first ten rows of the data set:
The data set has independent variables as several physiological parameters of a diabetes patient. The dependent variable is if the patient is suffering from diabetes or not. Here the dependent column contains binary variable 1 indicating the person is suffering from diabetes and 0 he is not a patient of diabetes.
dataset=pd.read_csv('diabetes.csv')
dataset.head()
# Printing data details
print(dataset.info) # for a quick view of the data
print(dataset.head) # printing first few rows of the data
dataset.tail # to show last few rows of the data
dataset.sample(10) # display a sample of 10 rows from the data
dataset.describe # printing summary statistics of the data
pd.isnull(dataset) # check for any null values in the data
Creating variables
As we can see that the data frame contains nine variables in nine columns. The first eight columns contain the independent variables. These are some physiological variables having a correlation with diabetes symptoms. The ninth column shows if the patient is diabetic or not. So, here the x stores the independent variables and y stores the dependent variable diabetes count.
x=dataset.iloc[:,:-1].values
y=dataset.iloc[:,-1].values
Performing the classification
To do the classification we need to import the DecisionTreeClassifier() from sklearn. This special classifier is capable of classifying binary variable i.e. variable with only two classes as well as multiclass variables.
# Use of the classifier
from sklearn import tree
clf = tree.DecisionTreeClassifier()
clf = clf.fit(x, y)
Plotting the tree
Now as the model is ready we can create the tree. The below line will create the tree.
tree.plot_tree()clf
Generally the plot thus created, is of very low resolution and gets distorted while using as image. One solution of this problem is to print it in pdf format, thus the resolution gets maintained.
# The dicision tree creation
tree.plot_tree(clf)
plt.savefig('DT.pdf')
Another way to print a high resolution and quality image of the tree is to use Graphviz format importing export_graphviz() from tree.
# Creating better graph
import graphviz
dot_data = tree.export_graphviz(clf, out_file=None)
graph = graphviz.Source(dot_data)
graph.render("diabetes")
The tree represents the logic of classification in a very simple way. We can easily understand how the data has been classified and the steps to achieve that.
Performing regression using decision tree
About the data set
The dataset I have used here for demonstration purpose is from https://www.kaggle.com. The dataset contains the height and weight of persons and a column with their genders. The original dataset has more than thousands of rows, but for this regression purpose, I have used only the first 50 rows containing data on 25 male and 25 females.
Importing libraries
Additional to the basic libraries we imported in a classification problem, here we will need to import the DecisionTreeRegressor() from sklearn.
# Import the necessary modules and libraries
import numpy as np
from sklearn.tree import DecisionTreeRegressor
import matplotlib.pyplot as plt
Reading the dataset
I have already mentioned about the dataset used here for demonstration purpose. The below code is to import the data and store in a dataframe called dataset.
dataset=pd.read_csv('weight-height.csv')
print(dataset)
Here is a glimpse of the dataset
Creating variables
As we can see that the dataframe contains three variables in three columns. The last two columns are only of our interest. We want to regress the weight of a person using the height of him/her. So, here the independent variable height is x and the dependent variable weight is y.
x=dataset.iloc[:,1:2].values
y=dataset.iloc[:,-1].values
Splitting the dataset
This is a common practice of splitting the whole data set for creating training and testing data set. Here we have set the test_size as 20% that means the training data set will consist 80% of the total data. The test data set works as an independent data set when need to test the classifier after it gets trained with training data.
# Splitting the data for training and testing
from sklearn.model_selection import train_test_split
x_train, x_test, y_train, y_test=train_test_split(x,y, test_size=0.20, random_state=0)
Fitting the decision tree regression
We have here fitted decision tree regression with two different depth values two draw a comparison between them.
# Creating regression models with two different depths
regr_1 = DecisionTreeRegressor(max_depth=2)
regr_2 = DecisionTreeRegressor(max_depth=5)
regr_1.fit(x_train, y_train)
regr_2.fit(x_train, y_train)
Prediction
The below line of codes will give predictions from both the regression models with two different depth values using a new independent variable set X_test.
# Making prediction
X_test = np.arange(50,75, 0.5)[:, np.newaxis]
y_1 = regr_1.predict(X_test)
y_2 = regr_2.predict(X_test)
Visualizing prediction performance
The below line of codes will generate a height vs weight scattered plot alongwith two prediction lines created from two different regression models.
# Plot the results
plt.figure()
plt.scatter(x, y, s=20, edgecolor="black",
c="darkorange", label="data")
plt.plot(X_test, y_1, color="cornflowerblue",
label="max_depth=2", linewidth=2)
plt.plot(X_test, y_2, color="yellowgreen", label="max_depth=5", linewidth=2)
plt.xlabel("Height")
plt.ylabel("Weight")
plt.title("Decision Tree Regression")
plt.legend()
plt.show()
Conclusion
In this post, you have learned about the decision tree and how it can be applied for classification as well as regression problem using scikit-learn of python.
The decision tree is a popular supervised machine learning algorithm and frequently used by data scientists. Its simple logic and easy algorithm are the main reason behind its popularity. Being a white box type algorithm, we can clearly understand how it is doing its work.
The DecisionTreeClassifier() and DecisionTreeRegressor() of scikit-learn are two very useful functions for applying decision tree and I hope you are confident about their use after reading this article.
If you have any question regarding this article or any confusion about its application in python post them in the comment below and I will try my best to answer them.