Improved Parallel Pattern Growth Data Mining Algorithm
P. R. Asha, T. Jebarajan · International Review on Computers and Software (IRECOS) · 2014
Data mining techniques that extract information from huge amount of data have become popular in many applications. Many algorithms are designed to analyze those volumes of data automatically in efficient ways. To improve the performance a data mining task, it is important that parallelism would be better than the sequential mining. Association Rule Mining (ARM) is data mining technique which aims to discover patterns/rules among items in a large database of variable length transactions. This paper proposes a parallel Frequent Pattern Tree Growth algorithm. Task parallelization is done by partitioning the database and sent to all of its compute nodes and finally the results were merged together in the Head node. Efficient partitioning and parallelization works in a better way and shows good performance. Filtering of the retrieved association rules using various Rule Interestingness measures has been done. The Performance of the parallel FP Tree algorithm is then compared and analyzed with the Rapid Miner Toolkit