Inductive Logic Programming for Knowledge Discovery in Databases

Stefan Wrobel. In Saso Dzeroski and Nada Lavrac, editors, Relational Data Mining, pages 74--101. Springer-Verlag, September 2001. More behind this link.

Abstract

Relational data mining algorithms and systems are capable of directly dealing with multiple tables or relations as they are found in today's relational databases. This reduces the need for manual preprocessing and allows problems to be treated that cannot be handled easily with standard single-table methods. This paper provides a tutorial-style introduction to the topic, beginning with a detailed explanation of why and where one might be interested in relational analysis. We then present the basics of Inductive Logic Programming (ILP), the scientific field where relational methods are primarily studied. After illustrating the workings of MIDOS, a relational methods for subgroup discovery, in more detail, we show how to use relational methods in one of the current data mining systems.

BibTeX entry.

Other publications

by S Wrobel,
of 2001.