基于传感器网络的分布式遗传k-means聚类算法研究

Research on Distributed Genetic K-means Clustering Algorithm Based on Wireless Sensor Networks

传统的集中式聚类算法不适宜对传感器网络的分布式数据进行聚类,用遗传进化机制对传统k-means的分布式聚类算法进行优化,可得出遗传k-means聚类算法。遗传k-means聚类算法即在传感器网络中sink节点传送随机选取的初始k个簇心到各个传感器节点,在这些节点上分别用遗传k-means聚类算法将本地的数据划分到距离最近的簇,然后将簇信息在无线传感器网络里通过路由逐层上传合并汇聚到sink节点,计算k个簇心的平均值,再往下传送k个簇心,反复迭代更新直至聚类目标函数值达到最小为止。实验表明,遗传k-means聚类算法的聚类效果较好,收敛速度较快。

The article focuses on research on clustering analysis of numerous high-dimensional redundant monitoring data in wireless sensor networks,and in light of inappropriate use of traditional centralized clustering algorithm in wireless sensor networks,The arlicle presents a distributed clustering algorithm,which uses the genetic evolution mechanism to optimize the traditional k-means clustering algorithm,resulting in genetic k-means clustering algorithm.The experiment of the algorithm shows the preferable clustering effect and fast convergence rate.