A LINQ-based Conditional Pattern Collection Algorithm for Parallel Frequent Itemset Mining on a Multi-Core Computer
Chun-Hong Huang, Yungho Leu · 2015
Due to the prevalence of big datasets, the existing frequent itemset mining algorithms are not adequate. In this paper, we propose a new algorithm for parallel frequent itemset mining on a multi-core computer system. We propose to use LINQ queries to divide a transaction database into a set of independent sub-datasets, called the conditional pattern collections. Subsequently, the threads of a multi-core computer system mine the conditional pattern collections concurrently by using an existing implementation of a frequent itemset mining algorithm. The experimental results showed that the proposed algorithm offered 2x speedup and 4x speedup, for a fast implementation of éclat and a fast implementation of FP-growth, respectively, on a personal computer system with a quad-core Intel i7 CPU.