Depot Institutionnel de l'UMBB >
Publications Scientifiques >
Communications Internationales >
Veuillez utiliser cette adresse pour citer ce document :
http://dlibrary.univ-boumerdes.dz:8080/handle/123456789/2077
|
Titre: | Fuzzy clustering for finding fuzzy partitions of many-valued attribute domains in a concept analysis perspective |
Auteur(s): | Djouadi, Y. Alouane, Basma Prade, H. |
Mots-clés: | Association rules Fuzzy C-means Fuzzy partitions Many-valued formal contexts Association rules mining Concept analysis Domain experts Formal contexts Fuzzy C mean Fuzzy C-means algorithms Human expert Information loss Knowledge discovery process Pre-processing Quantitative attributes Copying Fuzzy clustering Fuzzy logic Fuzzy systems Data mining |
Date de publication: | 2009 |
Référence bibliographique: | Joint 2009 International Fuzzy Systems Association World Congress, IFSA 2009 and 2009 European Society of Fuzzy Logic and Technology Conference, EUSFLAT 2009; Lisbon; Portugal; 20 July 2009 through 24 July 2009; Code 94760 |
Collection/Numéro: | 2009 International Fuzzy Systems Association World Congress and 2009 European Society for Fuzzy Logic and Technology Conference, IFSA-EUSFLAT 2009 - Proceedings 2009;pp. 420-425 |
Résumé: | Although an overall knowledge discovery process consists of a distinct pre-processing stage followed by the data mining step, it seems that existing formal concept analysis (FCA) and association rules mining (ARM) approaches, dealing with many-valued contexts, mainly focus on the data mining stage. An "intelligent" pre-processing of input contexts is often absent in existing FCA/ARM approaches, leading to an unavoidable information loss. Usually, many-valued attribute domains need to be first fuzzily partitioned. However, it is unrealistic that the most appropriate fuzzy partitions can be provided by domain experts. In this paper, an unsupervised learning stage, based on Fuzzy C-Means algorithm, is proposed in order to get fuzzy partitions that are faithful to data for quantitative attribute domains, and consequently for avoiding the loss of valuable association rules due to the use of empirical fuzzy partitions. More precisely, the paper reports an experiment where it is shown that some rules are no longer found because their support or confidence is too low when using such empirical partitions. Experimental results show that the learned fuzzy partition outperforms human expert fuzzy partitions. More generally, the paper provide discussions about the handling of many-valued attributes in both fuzzy FCA and fuzzy ARM |
URI/URL: | http://dlibrary.univ-boumerdes.dz:8080123456789/2077 |
ISBN: | 978-989950796-8 |
Collection(s) : | Communications Internationales
|
Fichier(s) constituant ce document :
|
Tous les documents dans DSpace sont protégés par copyright, avec tous droits réservés.
|