Decision Tree Development

By Matthew Zelinsky

Abstract:

I have been working on a decision tree that is written in C. A decision tree can be thought of as a 'classifier.' It is used to classify instances (representations of some object or task) based on their values for various features.

Before a decision tree can be used to classify anything, it first has to be created. A decision tree is created with a training set of instances. An instance has a set of feature values, and a classification. The algorithm that creates the tree looks at the set of instances at a given node, chooses the 'best' feature and value for that feature, and sends instances that have a value less than the chosen value for the chosen feature to the node's left child, and the other instances to the node's right child. This process repeats at the children nodes until all the instances at a node have the same classification. A leaf/classification node is then created, having that classification value.