Autoregressive and Arima Pro-Integrated Moving Average Models for Network Traffic Forecasting

The predictability of network traffic is critical in optimizing future network architectures, where an accurate traffic prediction model ensures high-quality service. This research focuses on forecasting network traffic using ARIMA models, particularly addressing significant traffic characteristics such as long-range dependence (LRD), self-similarity, and multifractality across various time scales. Utilizing Internet of Things (IoT) data, this study developed and evaluated multiple ARIMA models based on performance metrics including the Akaike Information Criterion (AIC), Schwarz Bayesian Criterion (SBC), maximum likelihood, and standard error. The research highlights the model's efficacy in capturing linear dependencies within network traffic, while also acknowledging the limitations of ARIMA models in handling data burst characteristics. Consequently, it suggests the potential integration with GARCH models to improve prediction accuracy by incorporating time-varying volatility.