Prediction of VOIP Traffic in Real Mobile Networks

Abstract

Voice Over Internet Protocol(VOIP) live analysis is also very viable where it concerns the maximisation of what the network resources can provide as far as its utilisation and the service that the mobile communication systems can provide through the most efficient avenue are concerned. The given paper is dedicated to Quality of Service parameters prediction of real measurements in the LTE-Advanced mobile network. Parameters used are modelled as a multivariate time series that not only examines how the measures evolved over time but also the nature of interaction of the measures with each other. To determine the accuracy of the prediction, the study uses three prediction algorithms: Vector Auto Regressive (VAR), Random Forest (RF), and Extra Gradient Boost (XGBoost). The raw data is first transformed into a form suitable for supervised learning, enabling the learning models to learn and make predictions about future traffic occurrences. In addition, the statistical tests are copied, including the correlations between the Quality of Service parameters and how to define call performance using the parameters. It illustrates that the suggested methodology would yield credible predictions of network behaviour, which may be used to optimise the network in the future and also to control VoIP traffic.