Auteur principal			Co-auteurs(s)
Nom :	Dr Terence Etchells		Nom :
Courriel :	t.a.etchells@livjm.ac.uk		Courriel :
Institution ou compagnie:	Liverpool John Moores University		Nom :
Département:	Computing & Mathematical Sciences		Courriel :
Ville :	Liverpool		Nom :
État/Province :	Merseyside		Courriel :
Pays :	United Knigdom		Nom :
Type de présentation:	Conférence :  50 minutes.		Courriel :
Conférence et numéro :	Derive & TI-CAS ,    Numéro : D03		Horaire :   Local :	jeudi, 10h30   1350
Site Internet :
Titre de la communication :	Derive, Neural Networks and the discovery of rules in data
Résumé de la communication :
This presentation describes how Derive was used to extract rules from trained Neural Networks that successfully classify data. A brief overview of Neural Networks and the role of rule extraction will be followed by a description of the datasets that the rule extraction method was applied to; including data that classifies breast cancer tumours into benign or malignant. In particular the presentation will deal with the tools and programming strategies that needed to be developed in Derive so that the rule extraction algorithm and validation of the rules could be efficiently implemented. In addition, the presentation will show how the 2 and 3 D plotting facilities of Derive can be used to visualise the rules extracted from data via a neural network.