Forecasting Network Traffic: A Survey and Tutorial With Open-Source Comparative Evaluation

Abstract

This paper reviews the literature on network traffic prediction and serves as a tutorial on the topic. We analyze various methodologies, including autoregressive moving average models (ARMA, ARIMA, SARIMA) and artificial neural network approaches (RNN, LSTM, GRU, CNN). Each method is presented with comprehensive mathematical foundations, enabling readers to understand the operation of different techniques fully. To provide practical insights, we conduct numerical experiments using real datasets to compare the fitting quality and computational costs of the various approaches. Additionally, we make our code publicly available, allowing readers to access a diverse set of forecasting tools and use them as benchmarks for developing more advanced solutions.