A decision tree is a flowchart-like tree structure where an internal node represents a feature(or attribute), the branch represents a decision rule, and each leaf node represents the outcome. The topmost node in a decision tree is known as the root node.
Decision trees come under the supervised learning algorithms category. It is primarily used for regression and classification in machine learning models. It provides transparency by offering a single view of all traces and alternatives. Decision tree also assign specific values to problems and decisions, enabling better decision-making.
But do you know how a decision tree works? Most people use it because it’s easy and provides a graphical representation of the problem. In this article, we will look at the decision tree algorithm in detail.
How Does Decision Tree Work?
Another decision tree algorithm CART (Classification and Regression Tree) uses the Gini method to create split points.
Where pi is the probability that a tuple in D belongs to class Ci.
The Gini Index considers a binary split for each attribute. You can compute a weighted sum of the impurity of each partition.
Entropy is a measure of disorder or uncertainty and the goal of machine learning models and Data Scientists, in general, is to reduce uncertainty.
We simply subtract the entropy of Y given X from the entropy of just Y to calculate the reduction of uncertainty about Y given an additional piece of information X about Y. This is called Information Gain. The greater the reduction in this uncertainty, the more information is gained about Y from X.
Let’s see an example to train model with diabetes data using the above algorithm.
Please note that, We are going to use Pandas and Sklearn for training data and using existing dataset of diabetes from Kaggle.
Github code: https://gist.github.com/pranavbtc/56aba86c6fe989a1d63353e17cc98426 (To open this link scroll down at bottom.)
Interpretation and visualization is made easy when Decision trees are used.
Capturing Nonlinear patterns is easier.
Normalization of columns is not needed as negligible data preprocessing is required from the user.
Variable selection can be more efficiently done.
Feature engineering such as predicting missing values can be done very efficiently using this algorithm.
There are no assumptions about distribution because decision tree has a non-parametric nature.(Source)
Overfitting noisy data and sensitivity to noisy data is a con.
Nominal variation in data can result in different decision tree. To reduce this con bagging and boosting algorithms are used.
Before creating a decision tree it is suggested to balance out the dataset as decision trees are biased to imbalanced dataset.
Decision tree is very easy to understand and communicate. It provides an excellent visual illustration of the data and the dendrogram gives a good look at the relationship between objects.