Time-Series Forecasting with SARIMAX for Intent Prediction

Modern network automation relies heavily on intent prediction, which allows systems to proactively manage resource utilization and predict user requirements over time scales. The application of time-series forecasting techniques, specifically the SARIMAX model, to intent prediction at various temporal scales in intent-based networking systems is examined in this paper. To comprehend complicated temporal behaviors and external determinants at short-, medium-, and long-horizon intervals, SARIMAX combines exogenous components with autoregressive, moving average, differencing, and seasonal variables. We provide a thorough examination of the ARMA family of models, emphasizing how well they can anticipate intent over a range of time periods and demonstrating how prediction accuracy has increased over time. The results we obtained show that intent-based designs and multiscale predictive capability can be combined to increase network agility, support dynamic applications like 5 G slicing, and allow for fine-grained QoS modifications for per-flow requirements in response to shifting temporal demands.