Please use this identifier to cite or link to this item:
Auteur(s): MOUNA, Azzeddine
Mots-clés: High performance Data mining
distributed association rules
Grids-based frequent
Itemsets mining
Data Grid
Date de publication: 2013
Résumé: DataMining techniques especially association rules allow to discover knowledges which help decision makers. We proposed a new strategy for the problem of distributed association rules on Grids particularly on the frequent Itemset mining step. Our approach consist in the modification of generation-pruning candidates Itemsets stage by introducing a new method based on the use of deductions on Itemsets support values for the frequent Itemsets mining and on the proposition of strategies more scalable and more suitable to the use of our method on general distributed frameworks and on Grids. Our approach allows on one hand to reduce the candidate Itemsets number and/or the database scans number and on the other hand to reduce the communications/synchronizations cost required for the exchange of this candidate Itemsets and/or for the calculation of the locales counts of these candidates in the different geographically distributed sites. The experiments made have allowed us to validate our approach and to prove it usefulness on improving the performances of the frequent Itemsets mining step on distributed contexts.
Appears in Collections:Thèses magister

Files in This Item:
File Description SizeFormat 
Magister_MOUNA.pdf3,05 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.