改进的Peer-to-Peer环境下的聚类算法

在P2PK-Means算法的基础上,提出了一种改进的数据聚类算法DK-Means。该算法不需要所有节点进行全局同步,只需要在直接相连的节点间进行通信,同时利用本地保存的直接相邻节点聚类信息来减少节点间的通信次数,从而减少整个网络的通信开销。与P2PK-Means算法的实验结果对比表明,改进后的算法通信量要小于P2PK-Means算法的通信量,并且在聚类准确度方面也没有损失,此外,随着节点的增多,DK-Means算法所需通信量的增长速度要明显低于P2PK-Means算法。

Improved clustering algorithm in peer-to-peer environments

P2PK-Means algorithm is a distributed clustering algorithm in Peer-to-Peer(P2P) network. An improved data clustering algorithm,DK-Means,was proposed based on P2PK-Means algorithm. The proposed algorithm works in a localized asynchronous manner by communicating with the directly connected nodes without global synchronization.DK-Means decreases communication overhead of P2P network by saving cluster information of direct neighbors,thus reducing the total network communication cost.Compared with P2PK-Means,the traffic is much less using the proposed DK-Means and there is no degradation in accuracy.Furthermore,with the increase in nodes,the traffic growth rate of DK-Means is lower that of P2PK-Means.