Rule Induction

Abstract

Rule induction has been researched for some decades within the larger field of Machine Learning. Machine Learning in its turn is a part of the Artificial Intelligence (AI) discipline which achieved recognition as a discipline in the early 50’s. The AI objective is to understand human intelligence and to develop intelligent systems. Machine Learning (ML) focuses on the ability of learning and gained momentum in the early 80’s with rule induction (also known as concept learning) for which it is still well known. Early successful applications of Machine Learning include: discovering rules of chemistry using Meta-Dendral [69], dis covering laws of physics using Bacon [260] and soybean disease diagnosis using AQU [298]. Apart from rule induction, other popular paradigms of the Machine Learning field are neural nets, genetic algorithms, case-based learning and analytic learning (theorem proving). Some early tutorials of the Machine Learning field can be found in [296,297], more recent overviews can be found in [121,210, 261,377,378]. Nowadays, several distinct reasons for using and studying Machine Learning can be observed:

• understanding human learning,

• developing computational learning,

• solving decision-making problems (e.g. classification),

• acquiring knowledge for expert systems,

• discovery of knowledge (data mining).