On Network Traffic Forecasting using Autoregressive Models

Various statistical analysis methods are studied for years to extract accurate trends of network traffic and predict the future load mainly to allocate required resources. Besides, many stochastic modeling techniques are offered to represent fundamental characteristics of different types of network traffic. In this study, we analyze autoregressive traffic forecasting techniques considering their popularity and wide-use in the domain. In comparison to similar works, we present important traffic characteristics and discussions from the literature to create a self-consistent guidance along with the survey. Then, we approach to techniques in the literature revealing which network characteristics they can capture offering a characteristic-based framework. Most importantly, we aim to fill the gap between the statistical analysis of those methods and their relevance with networking by discussing significant aspects and requirements for accurate forecasting from a network-telemetric perspective.