UNIVERSITY AT BUFFALO, THE STATE UNIVERSITY OF NEW YORK
The Department of Computer Science & Engineering

STUART C. SHAPIRO: CSE 116 B

Readings

Riley Chapter 9 ("classification trees")

Introduction

A decision tree is a tree with the following properties:

• each branch node represents a test to be performed;
• each branch node has one sub-decision tree for each possible result of its test;
• each leaf represents a final conclusion of the series of tests represented by the branch nodes on its branch from the root.

Decision trees are often used for learning classifications or categories.

The tests are often binary, yes/no or true/false, tests, but needn't be.

Example: A program to play the "Animal game", a version of twenty questions.

The program tries to guess the animal the user is thinking of. Each branch node is a yes/no question about the animal. Each leaf node is an animal.
If the program doesn't guess the animal, it solicits a question and further information from the user to use to grow the tree.
When the user quits playing, the program saves the expanded tree for the next user.

See file://localhost/projects/CSE116/ShapiroCode/Animals/AnimalTree.java
which you may run by executing:

cd /projects/CSE116/ShapiroCode/Animals/
animals &

Stuart C. Shapiro <shapiro@cse.buffalo.edu>