A decision tree is a classification approach that uses supervised learning.
This type of algorithm creates trees that anticipate the outcome of an input vector based on decision rules learned from the data’s attributes.
Decision trees are beneficial because they can visualize and grasp the circumstances that contribute to a specific outcome.
There are two different decision tree models: classification trees, which have a discrete goal variable and leaves that reflect class labels, and regression trees, which have a continuous target variable.
A set of data is used to train the tree, creating a model based on the data. The tree can then make decisions based on unknown facts.
There are numerous decision tree learning methods available. The Iterative Dichotomiser 3 (ID3) was one of the first, and it operates by separating a data set into two independent data sets based on a single field in the vector.
This field is chosen by calculating its entropy (a measure of the distribution of the values for that field). The goal is to pick a domain from the vector that will reduce entropy in successive data set splits as the tree grows.
Beyond ID3, C4.5 (a successor to ID3) and MARS (Multivariate Adaptive Regression Splines) build decision trees with better numerical data handling.