基于LTE小区评估的SOM和k-means算法比较

Comparison of SOM and k-means Algorithm Based on LTE Cell Evaluation

针对移动网络技术日益复杂的发展,为辅助网络运营商实现对移动网络的有效管理,提高对网络数据评估和优化的能力,通过无监督学习技术,提出一种基于关键性能指标(KPI)的特征提取法,对LTE小区进行模式聚类。通过对不同维度LTE小区样本数据特征的分析,对比自组织(SOM)神经网络和k-means算法的聚类表现,验证两种无监督学习聚类算法之间的优缺点。仿真结果表明,k-means与SOM在具有低维、高维数据集的不同小区中存在显著差异,分析实验数据总结出其中规律性。

In view of the increasingly complex development of mobile network technology,in order to assist network operators to realize effective management of mobile networks and improve their ability to evaluate and optimize network data,a feature extraction method based on KPI is proposed by unsupervised learning technology to cluster LTE cells.By analyzing the characteristics of LTE cell sample data with different dimensions,and comparing the clustering performance of SOM neural network and k-means algorithm,the advantages and disadvantages of the two unsupervised learning clustering algorithms are verified.The simulation results show that there are significant differences between k-means and SOM in different communities with low-dimensional and high-dimensional data sets,and the regularity is summarized by analyzing the experimental data.