Static Load Balancing of Parallel Mining of Frequent Itemsets Using Reservoir Sampling

0368102 - ÚI 2012 RIV DE eng C - Conference Paper (international conference)
Kessl, Robert
Static Load Balancing of Parallel Mining of Frequent Itemsets Using Reservoir Sampling.
Machine Learning and Data Mining in Pattern Recognition. Berlin: Springer, 2011 - (Perner, P.), s. 553-567. Lecture Notes in Artificial Intelligence, 6871. ISBN 978-3-642-23198-8. ISSN 0302-9743.
[MLDM 2011. International Conference /7./. New York (US), 30.08.2011-03.09.2011]
R&D Projects: GA ČR GAP202/10/1333
Institutional research plan: CEZ:AV0Z10300504
Keywords : frequent itemset mining * parallel algorithms * association rules * approximate counting
Subject RIV: IN - Informatics, Computer Science

In this paper, we present a novel method for parallelization of an arbitrary depth-first search (DFS in short) algorithm for mining of all FIs. The method is based on the so called reservoir sampling algorithm. The reservoir sampling algorithm in combination with an arbitrary DFS mining algorithm executed on a database sample takes an uniformly but not independently distributed sample of all FIs using the reservoir sampling. The sample is then used for static load-balancing of the computational load of a DFS algorithm for mining of all FIs.
Permanent Link: http://hdl.handle.net/11104/0202549