Practical Applications of Data Mining

100 Chapter 3 ClassifiCation learning

Then, it will be compared with its successors, which are modified versions

with additional features.

3.4.1 ID3

The underlying strategy of ID3 is to develop a decision tree recursively

from top to bottom by using the information gain measure of each attribute

as the criterion to select the attribute to test when splitting the dataset into

subsets.

As previously discussed, the leaves in the decision tree represent the

classes, and the internal nodes represent the attribute-based tests, which

are connected with the branches that represent the outcomes of each

test. Only one test is used for each branch. The process of identifying the

class of an instance starts from the root of