What are Decisions tree in Data Science?

This article will take a look into ‘What are Decisions tree in Data Science’ providing full understanding of the concept, with real world example and use cases.

The trees play an imperative part in human life, tree-based algorithms are a significant part of data science and machine learning.

The design of a tree has set the encouragement to build the algorithms and deliver them to the machines to analyze what we need them to learn and resolve complications in real life.

These trees based learning and analyzing algorithms are measured to be one of the finest and widely used techniques in supervised machine learning. Techniques like decision trees or random forests are being commonly used in all categories of problems of data science.

That is why, for every beginner in machine learning, it is very essential to learn these types of algorithms. So, in this article, I will explain “What are Decisions tree in Data Science”.

What are Decisions tree in Data Science | Basic Concepts

The decision tree method is a frequently used data mining technique for the establishment of classification schemes centered on various developing forecast algorithms for a target variable.

This technique categorizes a population into branches like nodes that build a tree with the main root node, internal nodes, branches, and leaf nodes.

This technique proficiently deals with large and complex datasets without appointing a difficult parametric arrangement. When the model size is large enough, given data can be distributed into the training dataset and validation dataset.

By means of the training dataset to form a decision tree classical and a validation dataset to conclude on the suitable tree size required to attain the ideal ultimate classical.

As the name recommends, a decision tree is a stream similar to a tree structure that chores on the fundamental of conditions. A decision tree is effective and has solid algorithms castoff for forecasting analysis.

The core modules and essential steps in constructing a decision tree classical are following;


There are mainly three kinds of nodes.

Source: Micro Medium

i)                    Root Node

A root node is also known as a decision node. It signifies an option that will outcome in the subsection of all data records into two or more jointly select subsets. This highest node also symbolizes the final outcome or final decision you are going to mark.

ii)                  Internal nodes

An internal node is also known as chance nodes. It signifies one of the probable selections presented at that edge in the tree structure. The top edge of that node is linked to its parent or its root node and the bottom edge is linked to its leaf nodes or child nodes.

iii)                 Leaf Nodes

A leaf node is also known as end nodes. The leaf nodes that are bound at the end of the branches symbolize probable results for each act. There are normally two kinds of leaf nodes: one is square leaf nodes that specify another decision to be prepared, second is circle leaf nodes that specify a chance occurrence or unknown conclusion.


Branches, that trunk from the root, symbolize diverse choices or sequences of action that are accessible when building a specific decision. Branches are most frequently specified with an arrow line.

A decision tree classical is shaped by means of a hierarchy of branches. Each track from the root node from side to side internal nodes to an end node or leaf node indicates a classification decision principle.


Splitting is a procedure of distributing a node into two or more than two sub-nodes. Only input datasets associated with the target dataset are castoff to divide parent nodes into their child nodes.

When constructing the classical, one must first recognize the most significant input dataset, and then divide the given records, at the root node, into two or more groups. This splitting process remains until pre-planned sameness or ending conditions come across.


Difficulty and complexity are competing for features of classical that require to be concurrently measured when constructing a classical. The more complex a classical is, the less trustworthy it will be when castoff to forecast future outcomes. To avoid this situation, stopping rules must be realistic when constructing a decision tree to avoid the classical from becoming excessively complex.

Mutual factors used in stopping rules comprise:

  1. The least possible number of data in an end node or leaf node
  2. The least possible number of data in a node preceding splitting
  3. The number of paces of any leaf or end node from the root node.


 In some circumstances, stopping rules is unable to work fine. Another method to construct a decision tree classical is to develop a big tree foremost and then prune that one to the best mass by eliminating nodes that deliver fewer information. Pruning is a frequently used process of selecting the best probable sub-tree from numerous trees. You may say that pruning is the opposite method of splitting.

What are Decisions tree in Data Science | How it Works

The decision of creating planned splits has a deep impact on the accuracy of a tree. The decision principles are different for regression and classification trees.

Decision trees practice many algorithms to adopt to split a node into sub-nodes. The formation of sub-nodes rises the sameness of resulting sub-nodes. You can say that the clarity of the node rises with respect to the target outcome.

The decision tree divides the nodes on all accessible datasets and then chooses the split which outcomes in most similar sub-nodes.

The plain indication behind a decision tree algorithm is as follows;

  • Choose the best feature using feature selection measures (FSM) to divide the records.
  • The sort that features a decision node and breakdowns the dataset into smaller subsections.
  • Starts tree structure by reiterating this procedure recursively for every child node until one of the desired condition will match:
  • All the tuples are in the right place to a similar feature value.
  • There are no left behind attributes.

Feature selection measures make available a rank to every feature by the explanation of the particular dataset. The top score feature will be nominated as a splitting feature.

Source: Wixstatic

What are Decisions tree in Data Science | Categories

Category of decision tree be determined by upon the type of input dataset which we have we have that is categorical or numerical:

Categorical decision tree.

If the given input dataset is a categorical type, for example, whether the loan challenger will nonpayer or not, its answer is either yes or no. This category of the decision tree is known as a Categorical decision tree.

Continuous decision tree.

If the given input dataset is numeric types and or is continuous, for example, when we have to forecast a plot price. Then the castoff decision tree is known as a Continuous decision tree.

What are Decisions tree in Data Science | Algorithms

The most famous varieties of decision tree algorithms are following;

1-     Iterative Dichotomiser 3 (ID3)

ID3algorithm practices Information obtain to conclude which characteristic is to be castoff categorize the present subsection of the data. For each phase of the tree, information obtained is calculated for the left behind data repetitively.

2-     C4.5

C4.5 algorithm is the inheritor of the ID3 algorithm. This algorithm practices either Information obtain or obtain ratio to conclude upon the categorizing characteristic. It is a straight development from the ID3 algorithm as it can manage both missing and continuous characteristic values.

3-     Classification and Regression Tree (CART):

Classification and Regression Tree (CART) is an active and energetic learning algorithm that be able to build a regression tree model as well as a classification tree model.

What are Decisions tree in Data Science | Real time Example

You’ve perhaps castoff a decision tree in the past in your life to mark a decision. Let’s proceeds with a simple example of the decision. Suppose you want to play a game “tennis” on a specific day with your son.

It may be contingent on different aspects similar to whether or not you will get free from your work on time and whether you arrive at your home before 10 am depending upon road traffic or whether your son is busy in some other doings already planned that day.

In all the circumstances, your decision for playing tennis with your son generally be determined by your availability and your son’s availability at that specific time and the weather condition, outside.

If the weather is good, and you have no work from the office and arrive at your home in time; your son does not have any work to do, you might desire to go out to the tennis play hall with him. If you arrive, on time, at home but your son already has scheduled some other work that daytime, then you cannot go to play tennis.

Source: Wixstatic

This is a perfect example of a real life and everyday decision tree. We have constructed a tree to model hierarchical decisions that finally describe some concluding results.

We have also selected our decisions to be fairly high-level to build the tree minor. For instance, what if we follow many probable choices for the weather like exact temperature (27 degrees sunny, 26 degrees raining 25 degrees sunny, 25 degrees raining,), etc., our tree will be massive.

The exact temperature actually is not much related, we just have to recognize whether it is fine to be outdoor or not.

So, we analyze all these aspects for the most recent past few days and build a table similar to the below.

source: Wixstatic

What are Decisions tree in Data Science | Advantages

Decision trees have been extensively used in many disciplines since these are very easy to be castoff, free of uncertainty, and strong even in the existence of missing values. Recently, the decision tree approach has become prevalent in the medical research field.

A sample of the medical usage of decision trees is in the diagnosis of a medical disorder from the configuration of indications, in which the classes, distinct by the decision tree, might either be dissimilar medical condition or subtypes, or patients with a situation who must take various therapies. However, decision trees have many advantages out of which some important are given below;

Easy to Understand

Decision trees outcomes are quite easy to understand and to recognize even for a person who is not from that background. These do not need any arithmetical and statistical knowledge to analyze and understand them. Its graphical demonstration is very simple and users can easily narrate their assumptions.

Useful in Data exploration

It is the quickest method to recognize the most important variables and relations among two or more variables. With the support of decision trees, we can make new features or variables that have well control to forecast target variables.

Less data cleaning required

It needs a smaller amount of data cleaning matched to other methods. It is not affected by missing values and outliers.

A Datatype is not a constraint

It is able to handle both categorical and numerical variables.

Non-Parametric Method

A decision tree is known to be a non-parametric technique. It means that decision trees have no hypothesizes about the classifier structure and the space distribution.

What are Decisions tree in Data Science | Disadvantages

  • Overfitting is one of the greatest applied trouble for decision tree classical. This complication gets resolved by implementing restrictions on classical parameters.
  • Decision tree reduces its information, while in work with a continuous numerical variable when it classifies variables in different classes.
  • Logics get distorted if even there are small variations in the training dataset.
  • Larger trees get problematic to adapt and explain.

What are Decisions tree in Data Science | Conclusion

In data science and machine learning, you cannot every time depend on linear models for the reason that there is non-linearity at extreme places. It is noted that decision trees treat in a moral mean with non-linearity. Decision tree techniques originate from supervised learning classical that could be castoff for both regression and classification jobs.

The job that is puzzling in the decision tree is to examine the aspects that choose the root node even though the outcomes in the decision trees are quite easy to understand.

Read More on Techniques used in Data:

What are the Techniques used in Data Science?

Applying Linear Regression techniques in Data Science

What are Clustering Techniques in Data Science?

About Post Author

Leave a Reply