Interactive Rule-network Layout With A Genetic Algorithm In A Knowledge Discovery Process
Price
Free (open access)
Volume
25
Pages
10
Published
2000
Size
1,175 kb
Paper DOI
10.2495/DATA000421
Copyright
WIT Press
Author(s)
P. Kuntz, R. Lehn & H. Briand
Abstract
In this paper, we adapt a recent graph drawing heuristic based on a Genetic Algorithm to some interaction constraints required in a human-centered process of Knowledge Discovery in Databases. The drawing problem is here a multi-objective problem: generating a clear layout of the extracted rules which satisfies readability requirements while minimizing disruptions of the "mental map" during the dynamical mining. We show that GA are particularly powerful for this problem. Experimental results are presented on both synthetic transaction data and random graphs. 1 Introduction If, as noticed in particular by Brachman and Anand 1996, most of the first existing systems developed in Knowledge Discovery in Databases (KDD) have been motivated mainly by new data-mining techniques rather than by the decision-maker's task, an increasing importance is give
Keywords