Result: COIL 2000 Challenge Solution based on ILLM-SG Methodology

Title:
COIL 2000 Challenge Solution based on ILLM-SG Methodology
Authors:
Tomislav Smuc Laboratory, Tomislav Šmuc
Contributors:
The Pennsylvania State University CiteSeerX Archives
Source:
http://www.liacs.nl/~putten/library/cc2000/SMUCPS~1.pdf.
Publication Year:
2000
Collection:
CiteSeerX
Subject Terms:
COIL 2000 Challenge, inductive learning algorithm, stacked generalization
Document Type:
Academic journal text
File Description:
application/pdf
Language:
English
Relation:
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.21.4385; http://www.liacs.nl/~putten/library/cc2000/SMUCPS~1.pdf
Availability:
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.21.4385
http://www.liacs.nl/~putten/library/cc2000/SMUCPS~1.pdf
Rights:
Metadata may be used without restrictions as long as the oai identifier remains attached to it.
Accession Number:
edsbas.47DB46F5
Database:
BASE

Further Information

This paper describes methodology used for solving tasks of COIL-2000 Challenge. Basic tools used in the detection task were ILLM algorithm (Inductive Learning by Logic Minimization) and stacked generalization. Induction of rules was performed in three steps. The first step involved optimization of parameters typical for ILLM induction process: noise level tolerance, maximum rule complexity. Optimized parameters were used in the final run and highest performance sub-rules were selected for stacked generalization. Description task involved explanation of sub-rules used in the detection task, and reconstructing 'customer models' from conjunction of attribute-value pairs. KEYWORDS: COIL 2000 Challenge, inductive learning algorithm, stacked generalization.