Scaling Up ILP to Large Examples: Results on Link Discovery for Counter-Terrorism (2003)
Inductive Logic Programming (ILP) has been shown to be a viable approach to many problems in multi-relational data mining (e.g. bioinformatics). Link discovery (LD) is an important task in data mining for counter-terrorism and is the focus of DARPA's program on Evidence Extraction and Link Discovery (EELD). Learning patterns for LD is a novel problem in relational data mining that is characterized by having an unprecedented number of background facts. As a result of the explosion in background facts, the efficiency of existing ILP algorithms becomes a serious limitation. This paper presents a new ILP algorithm that integrates top-down and bottom-up search in order to reduce search when processing large examples. Experimental results on EELD data confirm that it significantly improves efficiency over existing ILP methods.
In Proceedings of the KDD-2003 Workshop on Multi-Relational Data Mining (MRDM-2003), pp. 107--121, Washington DC, August 2003.

Prem Melville Ph.D. Alumni pmelvi [at] us ibm com
Raymond J. Mooney Faculty mooney [at] cs utexas edu
Lappoon R. Tang Ph.D. Alumni ltang [at] utb edu