top of page
Search

What is Decision Tree?, The way to make proper Decision

Introduction

It is popular tool for classification of choices and prediction of decisions. There was no clear history who invented it and how it was invented or when it was invented. In this technique you came across a diagram or chart that helps you to determine statistical probability or outcome of the choices you will made. It is mainly use to clarify, map out and find answer to complex problem. It has given a name because its resemblance like a inverted tree, a diagram which have root at top that branches out coming downward. A decision tree is a tree like structure where each upper node denotes the test on some subject and branch represent the possible outcome after passing through each test

Construction

A decision tree is drawn upside down with its root at the top. The construction of decision tree does not require any special knowledge. And so is easy to use and any one can make decision while drawing it. It gives you more accuracy in your decision that you have to make. It can also be use to learn knowledge which based on classification. Which can be used same as mind map.


Components

It has mainly three components

  • First is main point about which you have to make decisions. Called as MAIN NODE

  • Second is CHANCE OUTCOME, which show many ways if you choose a particular path to follow

  • And third and last is END POINT. Which is final outcome


  • Main Node

Firstly you have to set the main node or point about which you want to study the probabilities and have to make decision. At this node you have to make decision. Which can be displayed as square node with two or more arcs pointing to options. This arc represent possible outcomes

  • Chance Outcome

This also known as chance node which shows probability or certain result. It is also called as branches. Represented by circle. In this stage you have to split the main set, into subsets based upon some attribute value. Here each branch or subset represent possible decision or outcome. After chance node, if you have to make another decision then make other square node in after the circle. In this way you can come across the decisions and its probabilities. If the result of a decision node is another decision then keep dividing it.

  • Outcome

This also known as end node. At this node your diagram or map is going to end. It is also called as leaves. It is represented by triangle node. You have to keep splitting nodes up to when the splitting is no longer adds value to your prediction. Means triangle node in decision tree is final outcome. After which you cannot make decision it is final one.


So a path in decision tree from main square node to triangular node or leave show you possible decision which may show risk or reward. So it show the final and beneficial path you have to choose



Let take example you have to arrange a party then where you should arrange it? Indoor or outside in garden. This will be the decision you want to take, which can be shown as square node.

Then lets come to second point the chance outcomes, the chances here are there will be rain or there will be not rain in both cases, which can be shown by circular node. So we show them as chance outcomes or possibilities in our tree.

Then at last we come to the conclusion after the chance outcome, which can be shown as triangular node.

If we arrange it indoor and there will be rain then the party will get crowdy and dry there, and we are happy by proper feeling of being sensible. If we arrange party indoor and there will not be rain, then the party will be crowdy and and guest may feel hot. And we regret about what might better have been done. If the party is in garden and if there is rain, then it the worst condition. The rain will ruined the party, refreshment and damp guest. We feel unhappiness there. And at last if we arrange party outside and no rain is their, then there will be very pleasant party all guest will feel comfortable

How Decision Tree works

It is a graphical depiction of decision and its possible outcome or result. As it is graphical you can see whole picture which can be taken in different ways. It give us clear picture to identify every potential option and its outcome. By this we can see and predict what will happen in each path we follow. It shows us how and why one decision may lead to next, so we can be clear about what way to follow to be best. It helps us to understand the weight of each point you make. It allows multiple possible solution. And you can view each of them. It show the relationship between different decisions. So you can understand the interdepended of each decision. It helps to asses the benefit and drawbacks at end nodes. In decision tree the events must be mutually exclusive, means, if one event happens other cannot


Pruning

Pruning involves removing the branches that having low importance. By this way we can reduce the complexity of tree, and thus increasing its predictive power by reducing overfitting. It helps to keep the picture cleans and easily understandable. By pruning you can came to final decision very clearly and in short time.

Strength and Weaknesses of Decision Tree approach

Lets first see the strength

Able to generate understandable rules or paths. Which can be gain quickly in short period of time. Perform classification without requiring much computation. You don't have to make calculation and critical thinking. Provide a clear indication of which field are most important for prediction or classification. New possibilities can easily be added. If you think or identify another way or decision then add to this picture by making another branch. So it can be edited and properly moulded without making much changes. Little data preparation. You can draw large decision on single little piece of paper. Able to handle multi output problems. Because you are exposed to large decisional pathways and you come to know all the possible outcome


Weakness

Prone to errors in classification problems. But have low probability, which can be minimize if you think in clear way. Computationally expensive to train. Which make it difficult to make it in software. Still there are many application present on internet. Can be overly complex with too many pathways. If you make many different outcomes to each node. Over complication(Overfitting) lead to analysis paralysis. Means by seeing much more complication and making many branches to node you can't get to proper decision. This trees can be unstable because small variations in data might result in completely different tree being generated. Means if you make small changes near the root of the tree then all the way ahead of it get changes upto its final node.


Find more interesting videos here

Comments

bottom of page