基于小波分析的网络通信大数据谱聚类算法研究

Research on large data spectrum clustering algorithms for network communication based on wavelet analysis

传统谱聚类算法存在聚类效果差的缺陷,为此提出基于小波分析的网络通信大数据谱聚类算法研究。采用小波分析方法对采集到的电网通信网络大数据的相异性进行度量,将得到的相异性度量结果转换为数据之间的相似性,并对网络通信大数据相似性关系进行构建,得到网络通信大数据的相似度矩阵,以上述得到的网络通信大数据相似度矩阵为基础,采用聚类算法对数据进行聚类,实现了网络通信大数据的谱聚类。通过实验可得,提出的谱聚类算法的准确率与纯度分别高出传统算法34%与21.2%,说明提出的基于小波分析的谱聚类算法具备极好的聚类效果。

The traditional spectral clustering algorithm has the disadvantage of poor clustering effect,so a new spectral clustering algorithm for network communication large data based on wavelet analysis is proposed.Wavelet analysis method is used to measure the heterogeneity of big data collected from power network communication network.The results of the measurement of heterogeneity are converted into the similarity between the data.The similarity relation of the large data of network communication is constructed,and the similarity matrix of the large data of network communication is obtained.Based on the similarity matrix of the large data of network communication obtained above,clustering algorithm is used to calculate the similarity of the large data of network communication.Data clustering is used to realize spectral clustering of large data in network communication.The experimental results show that the accuracy and purity of the proposed spectral clustering algorithm are 34%and 21.2%higher than those of the traditional algorithm,respectively.It shows that the proposed spectral clustering algorithm based on wavelet analysis has excellent clustering effect.