Propositionalization and Redundancy Treatment

2nd International Workshop on Databases, Documents, and Information Fusion

Mark-A. Krogel ; Stefan Wrobel

Others  :  http://CEUR-WS.org/Vol-124/03krogel.pdf
PID  :  1660

来源: CEUR

【 摘 要 】

Following the success of relational learning/inductive logicprogramming on structurally complex but small problems, recently therehas been strong interest in relational methods that scale to real-worlddatabases (relational data mining). Transformation-based methods havealready been shown to be particularly promising approaches for robustlyand effectively handling larger relational data sets. However, these methods also pose problems: beside the cost of joins, they can produce verylarge numbers of features (attributes), and among those there are possibly higher proportions of redundant features. In this paper, we investigate the extent of redundancies and approaches to dealing with them.We present promising results from experiments in several domains.