Algorithm for Rule Induction

Instances: a set of training instances
Features: feature vector (f1, f2, ..., fn) input from teacher
Domains: set of domains for Features {d1, d2, ..., dm}
Classes: set of classes {c1, c2, ..., ck}
(simply {+, -} for single concept learning.)

Function formrule (Instances, Features, Domains, Classes)
For some class &isin Classes
If all members of Instances fall into class
then return class
else f &larr select-feature (Features, Instances)
d &larr domain from set Domains corresponding to f
return a tree of the form:

Contents    Page-10    Prev    Next    Page+10    Index