Publicação

Using deep neural networks for forecasting cell congestion on LTE networks: a simple approach

Ver documento

Detalhes bibliográficos
Resumo:Predicting short-term cellular load in LTE networks is of great importance for mobile operators as it assists in the efficient managing of network resources. Based on predicted behaviours, the network can be intended as a proactive system that enables reconfiguration when needed. Basically, it is the concept of self-organizing networks that ensures the requirements and the quality of service. This paper uses a dataset, provided by a mobile network operator, of collected downlink throughput samples from one cell in an area where cell congestion usually occurs and a Deep Neural Network (DNN) approach to perform short-term cell load forecasting. The results obtained indicate that DNN performs better results when compared to traditional approaches.
Autores principais:Torres, Pedro
Outros Autores:Marques, Hugo; Marques, Paulo; Rodriguez, Jonathan
Assunto:LTE SON Machine learning Deep learning Forecasting
Ano:2018
País:Portugal
Tipo de documento:capítulo de livro
Tipo de acesso:acesso restrito
Instituição associada:Instituto Politécnico de Castelo Branco
Idioma:inglês
Origem:Repositório Científico do Instituto Politécnico de Castelo Branco
Descrição
Resumo:Predicting short-term cellular load in LTE networks is of great importance for mobile operators as it assists in the efficient managing of network resources. Based on predicted behaviours, the network can be intended as a proactive system that enables reconfiguration when needed. Basically, it is the concept of self-organizing networks that ensures the requirements and the quality of service. This paper uses a dataset, provided by a mobile network operator, of collected downlink throughput samples from one cell in an area where cell congestion usually occurs and a Deep Neural Network (DNN) approach to perform short-term cell load forecasting. The results obtained indicate that DNN performs better results when compared to traditional approaches.