Machine Learning for OpenCV 4 - Second Edition

Investigating the inner workings of a decision tree

We established earlier that a decision tree is basically a flow chart that makes a series of decisions about the data. The process starts at the root node (which is the node at the very top), where we split the data into two groups (only for binary trees), based on some decision rule. Then, the process is repeated until all remaining samples have the same target label, at which point we have reached a leaf node.

In the spam filter example earlier, decisions were made by asking true/false questions. For example, we asked whether an email contained a certain word. If it did, we followed the edge labeled true and asked the next question. However, this works not just for categorical features, ...