Community Detection Applications in Mobile Networks: Towards Data-Driven Design for Next Generation Cellular Networks

The existing radio access network (RAN) is facing many challenges to meet the very strict speed and latency requirements by different mobile applications in addition to the increasing pressure to reduce operating cost. Innovation and development in RAN have been accelerated to tackle these challenges and to define how next generation mobile networks should look like. The role of machine learning (ML) and artificial intelligence (AI) driven innovations within the RAN domain is strengthening and attracting lots of attention to tackle many of the challenging problems. In this paper we surveyed RAN network base stations (BSs) clustering and its applications in the literature. The paper also demonstrates how to leverage community detection algorithms to understand underlying community structures within RAN. Tracking areas (TAs) novel framework was developed by adapting existing community detection algorithm to solve the problem of statically partitioning a set of BSs into TA according to mobility patterns. Finally, live network dataset in dense urban part of Cairo is used to assess how the developed framework is used to partition this part of the network more efficiently compared to other clustering techniques. Results obtained showed that the new methodology saved up to 34.6% of inter TA signaling overhead and surpassing other conventional clustering algorithms.