Analysis of Self-Organizing Maps (SOM) Methods for Cell Clustering with High-Dimensional OAM Collected Data

Mobile network produces a large amount of Operations, Administration and Maintenance (OAM) data. These data include different parameters of the base station and performance information of different cells. The Self-organizing Map (SOM) is an effective neural network for decreasing dimension, visualizing and clustering high dimensional data. It always transforming input vectors into one or two-dimensional grid prototype vectors and reorder them into different groups, which are easier to characterize and visualize the original data. In this work, our objective of this study is to explore and extract several representative cell patterns, at the same time visualizing and clustering cells via the data set analysis by using Self-Organization Maps (SOM) methods and cluster cells into different groups based on various of features in radio access network (RAN) in LTE. We built a SOM model of the network using vectors from all LTE cells. Obtained results show the successful cluster in cells achievable by the SOM.