MEC环境下多维属性感知的边缘服务二次聚类方法研究

Secondary clustering method for edge services based on multi-dimensional attribute perception in MEC environment

移动边缘计算(mobile edge computing,MEC)环境下,海量的领域服务分布在边缘服务器上,如何对大规模的边缘服务进行精确的聚类是亟需解决的重要问题之一。为此提出了一种MEC环境下多维属性感知的边缘服务二次聚类方法。该方法首先分析并建立了MEC环境下边缘服务二次聚类指标模型。之后,提出了一种基于密度的最小生成树启发式分段聚类算法(heuristic segmented for MST clustering based on service density,DMSC),基于DMSC算法依据一级指标对边缘服务进行一次聚类。最后,将密度峰值算法中γ值引入到层次聚类中,构建了基于密度峰值的层次聚类算法(hierarchical clustering based on density peak,HCDP),基于HCDP算法依据二级聚类指标,在一次聚类的基础上对边缘服务进行二次聚类。在人工数据集和UCI数据集上开展了大量验证实验。实验结果表明,DMSC算法与HCDP算法提高了聚类的准确率,减少了算法的平均迭代次数,增强了算法的稳定性。

In MEC environment,many services distribute on edge servers.Accurate clustering of a large number of edge ser-vices in MEC environment is important.Therefore,this paper proposed a multi-dimensional attribute-aware edge service se-condary clustering method in MEC environment.Firstly,it analyzed and established the secondary clustering index model of edge service in MEC environment.After that,it proposed a heuristic segmented for MST clustering based on service density(DMSC) algorithm,and clustered edge services based on first-level index.Finally,it introduced the γ in the density peak clustering algorithm into the hierarchical clustering,and proposed a hierarchical clustering based on density peak(HCDP) algorithm.Experiments show that the DMSC algorithm and the HCDP algorithm improve the accuracy of clustering,reduce the average number of iterations of the algorithm.