CODEWITHSUNDEEP
artificial-intelligenceclassificationdecision-tree
cxcfan
cxcfan

Reputation: 185

Attribute selection for ID3 when there are more than 2 classes

Usually, we discuss attribute selection in ID3 with the highest information gain based on assumption that there are two classes: positive class and negative class. However, I just meet a problem where there are three classes. How to apply attribute selection algorithm for the case of three classes?

Upvotes: 1

Views: 440

Answers (1)

cxcfan
cxcfan

Reputation: 185

For convenience many books only discuss the calculation of I(p,n). If there are more than 2 classes, we can calculate I(c1,...,cn)=-C1*logC1-C2*logC2... while C1 is the proportion of elements of class 1.

Upvotes: 1

Related Questions