Parallel Optimized Algorithm for Apriori Association Rule Mining on Graphics Processing Unit with Compute Unified Device Architecture (CUDA)

Parallel computing is a form of computation in which many calculations are carried out simultaneously, operating on the principle that large problems can often be divided into smaller ones, which are then solved concurrently .Now GPU(Graphics Processor Unit) has taken a major role in high performance computing for general purpose applications. Compute Unified Device Architecture (CUDA) programming model provides the programmers adequate C language like APIs to better exploit the parallel power of the GPU. Data mining has significant applications in various domains. Currently, data mining toolkits cannot meet the requirement of applications with large scale database in terms of computation and speed. In this paper, we propose two parallel association rule mining algorithms i.e. parallel Apriori algorithm and parallel Apriori_Cluster algorithm on GPU with CUDA platform. These algorithms are implemented in CUDA with C language, which focuses on computation time compared with execution time of serial program in CPU.This parallel implementation out forms significantly on Tesla C2070 GPU and an Intel core i7 @ 3.40 GHz CPU. Our clustering of nodes is based on fiber optic communication. Our results have shown that GPU + CUDA parallel architecture is feasible for performance measurement in terms of execution time and promising for data mining applications. Keywords— Parallel computing, GPU, CUDA, Cluster, Data mining, Parallel Apriori Association rule mining