Self-Healing Framework for Next-Generation Networks through Dimensionality Reduction

Abstract

Next-generation self-organizing networks (NG-SONs) are the key that will lead to the full automation of the network management in the forthcoming generations of cellular communications. New challenges, like the deployment of novel wireless services or the aim of operators to provide end-to-end monitoring and optimization, make it necessary to develop an innovative scheme for network management. Within SON, self-healing (SH) comprises fault detection, root cause analysis (RCA), and compensation. Within these, the automation of RCA activities is one of the key elements to reduce operational expenditures related to network management. In this article, an SH framework for next-generation networks using dimensionality reduction is proposed as the tool enabling the management of an increasingly complex network, taking advantage of both feature selection and feature extraction techniques. A proof of concept has been carried out in the context of automatic RCA in a live network. Results show that the proposed framework can effectively manage a high-dimensional environment from different data sources, eventually automating the tasks usually performed by troubleshooting experts while optimizing the performance of the RCA tool.