Efficient implementation of apriori algorithm on HDFS using GPU

Abstract

A very efficient distributed processing framework is provided by Hadoop. For processing big data, Hadoop uses map-reduce programming model. The proposed technique uses parallel apriori mapreduce algorithm using high performance GPU. The computationally intensive operations of mapping phase are offloaded to GPU. Apriori is a very basic data mining algorithm which is used to determine the frequent item sets in the transactional database. In Hadoop, big transactional database are stored in structured form. When the size of transactional database is big, very fast apriori technique is required to solve the problem. Past researches show a clear view of solving data mining operations in heterogeneous environment which increase the performance with a very high rate than older serial execution techniques. This paper introduces integration of GPU in mapreduce programming model to solve the apriori data mining technique in a very time efficient manner. For our experimental implementation, we use NVIDIA's GPU and for the integration process, we use JCUDA and JNI.