Feature Extraction for Dimensionality Reduction in Cellular Networks Performance Analysis
Open Access
Abstract
Next-generation mobile communications networks will have to cope with an extraordinary amount and variety of network performance indicators, causing an increase in the storage needs of the network databases and the degradation of the management functions due to the high-dimensionality of every network observation. In this paper, different techniques for feature extraction are described and proposed as a means for reducing this high dimensionality, to be integrated as an intermediate stage between the monitoring of the network performance indicators and their usage in mobile networks’ management functions. Results using a dataset gathered from a live cellular network show the benefits of this approach, in terms both of storage savings and subsequent management function improvements.Keywords
Funding Information
- Ministerio de Economía y Competitividad (RTC-2017-6661-7)
This publication has 16 references indexed in Scilit:
- Autonomic Handover Management for Heterogeneous Networks in a Future Internet Context: A SurveyIEEE Communications Surveys & Tutorials, 2019
- Self-Healing Framework for Next-Generation Networks through Dimensionality ReductionIEEE Communications Magazine, 2018
- Automatic Feature Selection Technique for Next Generation Self-Organizing NetworksIEEE Communications Letters, 2018
- Unsupervised Technique for Automatic Selection of Performance Indicators in Self-Organizing NetworksIEEE Communications Letters, 2017
- Effective Feature Selection for 5G IM Applications Traffic ClassificationMobile Information Systems, 2017
- Data mining for fuzzy diagnosis systems in LTE networksExpert Systems with Applications, 2015
- Automatic Root Cause Analysis for LTE Networks Based on Unsupervised TechniquesIEEE Transactions on Vehicular Technology, 2015
- Diagnosis Based on Genetic Fuzzy Algorithms for LTE Self-HealingIEEE Transactions on Vehicular Technology, 2015
- Challenges in 5G: how to empower SON with big data for enabling 5GIEEE Network, 2014
- Independent component analysis, A new concept?Signal Processing, 1994