Rule LoE is a decision rule learning system. It is part of the LoE (League of Experts) framework. The main goal is to learn a set of rules which can be used to explain the predictions of a model. Rules are derived by combining a set of disjunct decision trees and estimating their competence area. Rule GoN can under some circumstances simplify (connect) rules of same type to simpler rules.
Having a set of n
decision trees (HDTrees supported only), the method will train one (or optionally more) decision tree which learns to approximate the competence area (the specialization) of the each tree.
This is a n
-class classification problem, where each class represents the competence area of a decision tree.
By then extending the path from this suorragate down to the leafs of the decision trees, a set of rules can be derived.
This package cannot be used standalone, but is part of a bigger framework called LoE (League of Experts). The main repository can be found here