Metropolitan Cellular Traffic Prediction Using Deep Learning Techniques

Abstract

With the advent of 5G networks, it is of paramount importance for machines to learn and make decisions independently. An important area where machine learning can be used to enhance wireless network performance is cellular traffic prediction. Cellular traffic volume prediction can be defined as forecasting the future traffic volume based on knowledge from the past, and other previously known information. This allows for congestion control and enhances energy efficiency, as the base station can be turned on and off based on incoming traffic data. In this research paper, a deep learning approach for cellular traffic prediction by using deep neural networks to model cellular traffic is proposed. This is achieved by treating the traffic volume data as a tensor, similar to an image, which is then fed to a convolutional neural network. The network learns the temporal and spatial dependence of cellular traffic data. The results of the proposed networks are then validated using the Telecom Italia Dataset.