A comparative approach of unsupervised machine learning techniques for LTE network parameter clustering

This paper presents a multivariate analysis of Long Term Evolution (LTE) cellular network physical layers parameters. Supporting a Multiple Input Multiple Output (MIMO) configuration three types of approaches were explored: partitional clustering through k-means algorithm, hierarchical clustering and density-based clustering through DBSCAN. Collected data were clustered with unsupervised blind machine learning methods. Techniques were applied on the same dataset depicting LTE radio layer quantities. Aim of the research is to find similarities among the implementation of the different cluster algorithms used to characterise the performance of the network.