Abstract

In recent years, data size has raised from Terabytes to Petabytes level. Traditional parallel algorithms for mining frequent patterns lack techniques in the new big data. The traditional algorithms cannot satisfy the requirements of big data analytics. Hence, it is necessary to develop new mechanism a parallel algorithm for mining frequent pattern over big data based on MapReduce. In this paper, new parallel mining of weighted frequent patterns using MapReduce is proposed and it is an alternative proposal over conventional frequent patterns method. The proposed method referred as Parallel mining of Weighted Frequent Pattern Tree using MapReduce (WFPTMP) increases the speed of the mining performance for high-dimensional data and the performance analysis is compared with the standard FP-Tree method. The experimental evaluation is performed using real-world data, which demonstrates that the proposed method is outperforming on execution time. Further, significant difference is identified on the selected statistical measures over existing algorithm.

Paper Details