Treatment learner
|
|
From [1]: In Data mining, treatment learners are used to find rules that change the expected class distribution (compared to some baseline).
Treatment learners include the J48, J48part, and APRIORI for discrete classes, TAR2 and TAR3 for weighted discrete classes, and the M5 for continuous classes. For an overview of these treatment learners see Tim Menzies' Data Mining Page (http://www.cs.pdx.edu/~timm/dm/kinds.html).
Classifiers are used for recognition tasks; e.g. find what toys to remove from the assembly line because they are defective.
Treatment learners are used for planning some minimal action to improve the odds that something will be later be recognized as belonging to some class. A treatment learner could be used to make repairs to the defective toys rejected from the assembly line.
Treatment learners are all about minimality - what is the least you need to do to most effect something.
See Also
External Sources
[1] T. Menzies. Introduction to Treatment Learning (http://www.cs.pdx.edu/~timm/dm/rx.html#treatment%20learning). Department of Computer Science, Portland State University.