Unified View of Decision Tree Learning Machines for the Purpose of Meta-learning

Abstract

The experience gained from thorough analysis of many decision tree (DT) induction algorithms, has resulted in a unified model for DT construction and reliable testing. The model has been designed and implemented within Intemi - a versatile environment for data mining. Its modular architecture facilitates construction of all the most popular algorithms by combining proper building blocks. Alternative components can be reliably compared by tests in the same environment. This is the start point for a manifold research in the area of DTs, which will bring advanced meta-learning algorithms providing new knowledge about DT induction and optimal DT models for many kinds of data.KeywordsDecision treesmeta-learningobject oriented design