The Application of Clustering Techniques for Anomaly Detection in 5G Access Networks
Clustering; Anomaly Detection; 5G Radio Access Networks.
The Internet arising and Information and Communication Technologies development expanded the volume and diversification of data sources, thus opening up new opportunities in the industry and academic fields for the application of Machine Learning and Big Data related technologies. Among the mentioned techniques, Clustering algorithms emerge as excellent solutions for exploratory data analysis and pattern recognition, mainly with the use of density based, distribution based and hierarchical clustering approaches. In the same perspective is the extensive amount of data generated by the Mobile Access Networks infrastructures. The Radio Access Networks, crucial for the telecommunications infrastructure, act as an enabler for the wireless communications and produce a significant volume of data coming from the collection of network counters, which, associated, allow monitoring and visibility regarding the network performance indicators and service quality. Through the usage of data analysis techniques, operators can obtain valuable insights towards network performance, user behavior and operational efficiency, leading to better network management and an improved user experience. The present work consists of applying clustering algorithms for cell segmentation through the usage of 5G Radio Access Network indicators data sets with features related to traffic, volume and channel quality, aiming to identify the cells with anomalous performance profiles - exposing unexpected user behavior changes or bad configuration - which consists of an essential activity to enable correspondence analysis between occurrences and applied configuration, discovery of resource allocation improvement opportunities and samples resource allocation for posterior supervised learning tasks.