Parallel Computing for Mining Association Rules in Distributed P2P Networks

Abstract

Distributed computing and Peer-to-Peer (P2P) systems have emerged as an active research field that combines techniques which cover networks, distributed computing, distributed database, and the various distributed applications. Distributed Computing and P2P systems realize information systems that scale to voluminous information on very large numbers of participating nodes. Data mining on large distributed databases is a very important research area. Recently, most work for mining association rules focused on a single machine or client-server network model. However, this traditional approach does not satisfy the requirements from the large distributed databases and applications in a P2P computing system. Two important challenges are raised, one is how to implement data mining for large distributed databases in P2P computing systems, and the other is how to develop parallel data mining algorithms and tools for the distributed P2P computing systems to improve the efficiency. In this chapter, a parallel association rule mining approach in a P2P computing system is designed and implemented, which satisfies the distribution of the P2P computing system well and makes parallel computing become true. The performance and comparison of the parallel algorithm with the sequential algorithm is analyzed and evaluated, which presents the parallel algorithm features consistent implementation, higher performance, and fine scalable ability.