Keywords: parallel processing, data mining, frequent itemsets, association rules, load balancing
Received May 25, 2009; revised August 24 & October 7, 2009; accepted November 18, 2009.
Communicated by Xiaodong Zhang.
| Previous | [ 1] | [ 2] | [ 3] | [ 4] | [ 5] | [ 6] | [ 7] | [ 8] | [ 9] | [ 10] | [ 11] | [ 12] | [ 13] | [ 14] | [ 15] | [ 16] | [ 17] | [ 18] | [ 19] | [ 20] | [ 21] | [ 22] | [ 23] | [ 24] | [ 25] |
¡@
S. M. FAKHRAHMAD1 AND GH. DASTGHAIBYFARD2
1Department of Computer Engineering
Islamic Azad University, Shiraz Branch
Shiraz, Iran
2Department of Computer Science and Engineering
Shiraz University
Shiraz, Iran
E-mail: {mfakhrahmad@cse.; dstghaib@}shirazu.ac.ir
One of the important and well-researched problems in data mining is mining association
rules from transactional databases, where each transaction consists of a set of
items. The main operation in this discovery process is computing the occurrence frequency
of the interesting set of items. i.e., Association Rule mining algorithms search for
the set of all subsets of items that frequently occur in many database transactions. In
practice, we are usually faced with large data warehouses, which contain a large number
of transactions and an exponentially large space of candidate itemsets, which have to be
verified. A potential solution to the computation complexity is to parallelize the mining
algorithm. In this paper, four parallel versions of a novel sequential mining algorithm for
discovery of frequent itemsets are proposed. The parallelized solutions are compared
analytically and experimentally, by considering some important factors, such as time
complexity, communication rate, load balancing, etc.
Received May 25, 2009; revised August 24 & October 7, 2009; accepted November 18, 2009.
Communicated by Xiaodong Zhang.