Machine Learning-based Approaches Comparison for Netflix/DAZN Streaming and Real Traffic Prediction

Traffic prediction is a fundamental topic in modern networks for performance analysis and network planning. To this aim, modeling and optimization of modern heterogeneous network infrastructures are key factors to achieve such goals, e.g. in terms of traffic flow improvement while reducing bandwidth allocation. Identifying an accurate model of a network device (e.g., a switch or a router) behavior is crucial in order to apply advanced and optimal resource allocation techniques. Such a problem is very challenging due to non-linearities and unavailability of internal variables measurements in real devices. In this respect, a promising direction is given by an appropriate integration of System Identification and Machine Learning (ML) techniques to obtain predictive models using historical data collected from the network. In this paper, a comparison of various ML-based methodologies to learn accurate models of the dynamical input-output behavior of a network switch device, obtained by appropriately combining AutoRegressive (AR) model identification with Regression Trees (RTs) and Random Forests (RFs), is provided. Furthermore, a validation over a real dataset obtained from measurements of an Italian Internet Service Provider (Sonicatel S.r.l.) is provided, as well as a comparison with classical and widely used ML methods.