基于热点度轨迹显影机制的网络社区聚类算法

为解决网络社区聚类算法在实际应用中存在热点捕捉困难和社区聚类生存时间较低的问题,提出了一种基于热点度轨迹显影机制的网络社区聚类算法。首先,考虑网络社区聚类存在的多径一体特性,采用抽样方式与角度估计方法来实现热点信号的精确捕捉,以提高聚类效率;随后,对热点信号矢量空间进行按列重排,并综合考虑传输矩阵具有的按列正交及全秩特性,构建热点度轨迹显影方法,以提高聚类中热点显影速度和增加聚类生存时间。仿真实验表明:与聚类流动性映射算法(Clustering Liquidity Mapping Algorithms,CLM算法)、超欧里几何热度聚类算法(Hyper-Eulerian Geometric Thermal Clustering Algorithms,H-EGTC算法)相比,所提算法具有更低的聚合时间和搜寻失误率,以及更高的热点显示时间。

THE CLUSTERING ALGORITHM OF NETWORK COMMUNITY BASED ON HOTSPOT TRAJECTORY DEVELOPMENT MECHANISM

In order to solve some shortcomings of the clustering algorithm in practice, such as difficulty in quickly catching hot spots, short clustering lifetime and high energy consumption, a clustering algorithm based on hotspot trajectory development mechanism is proposed. Firstly, considering the characteristics of multi-path integration which widely exists in network community clustering, and considering that the current network community clustering algorithms are mostly used in wireless sensor networks and the Internet of Things and other application scenarios, the sampling method and angle estimation method are used to achieve accurate capture of hot spot signals and enhance the efficiency of clustering aggregation process.Then, the hot spot signal vector space is rearranged according to the column for comprehensive consideration. Considering the orthogonal and full rank characteristics of the transfer matrix, a simple sorting mode of hotspot trajectory development method is constructed to improve the speed of hotspot development in clustering and increase the survival time of clustering. The simulation results show that compared with CLM and H-EGTC algorithms, the proposed algorithm has the advantages of less aggregation time, higher hotspot display time, lower search error rate and superior energy consumption. It has strong practical deployment value.