The growth in the number of services, functionalities and users expected with the arrival of the 5th Generation (5G) of mobile communications networks will produce an increase in the complexity of the network management tasks. These tasks should be adapted to meet the requirements of zero latency and great bandwidth presented by 5G new services. Thus, traditional management reactive approach should evolve to a proactive one. In this context, prediction techniques will become an essential tool. At the same time, 5G management algorithm will use both internal (network information) and external (context information) information sources. Thus, 5G management tasks should handle large amounts of data. The use of large dataset to generate prediction models can lead to the lack of precision of these models, due to the sparsity suffered by the data. In this paper, the use of feature selection techniques to improve prediction models is proposed.
Buscar en Dimension