ID3

ID3 (Iterative Dichotomiser 3) and its later variants including C4.5 of concept learning, developed by Quinlan, Ross. ID3 builds a decision tree in a top-down fashion using information measures, typically entropy, to choose a top level decision and the inductively use the same princple for lower branches. The simplest version of ID3 simply measures the entropy gain of the decision by counting the frequency examples of each class in the branches. As the name suggests, basic ID3 was developed for binary decisions, but there are modifications for non-binary decsions). There are also modifications to make the trees comprehensible (see expaliable AI), either by modifying the inductive learning algorithm itself or pruning the tree once constructed.

Defined on page 98

Used on pages 92, 98, 101, 102, 105, 183, 500, 502, 508