PRB-RUPFormer: A Recursive Unified Probabilistic Transformer for Residual PRB Forecasting

2026-05-14

AI Summary

Existing methods for predicting residual Physical Resource Blocks (PRBs) rely solely on single-carrier historical data, failing to capture cross-carrier dependencies and lacking uncertainty quantification—limitations that hinder effective spectrum decision-making in highly dynamic scenarios. This work proposes PRB-RUPFormer, the first unified probabilistic Transformer model tailored for residual PRB forecasting. By integrating temporal, seasonal, and carrier-aware embeddings, it jointly models multivariate KPI time-series couplings within a recursive multi-step prediction framework and outputs prediction intervals via quantile regression. The model employs a single shared architecture to efficiently learn network-wide traffic dynamics. Evaluated on six months of real-world LTE data, PRB-RUPFormer achieves median MAE below 0.05 and hit rates exceeding 0.80 in both one-day and seven-day recursive forecasts, significantly enabling intelligent RAN functionalities such as dynamic carrier activation and congestion avoidance.

Abstract

Accurate forecasting of residual Physical Resource Blocks (PRBs) is critical for proactive network slice provisioning, energy-efficient operation, and spectrum-aware decision making in cellular systems, where residual PRBs serve as a practical proxy for short- and medium-term spectrum availability. Existing PRB prediction methods typically rely only on historical PRB values and are trained independently per carrier or sector, limiting their ability to capture cross-carrier dependencies and providing no measure of forecast uncertainty. Moreover, point forecasts alone are insufficient for robust spectrum-aware control under highly variable traffic conditions. This paper proposes PRB-RUPFormer, a recursive unified probabilistic Transformer for residual PRB forecasting. The proposed model jointly processes multivariate KPI time series using temporal, seasonal, and carrier-aware embeddings, preserving inter-metric temporal coupling during recursive rollout and stabilizing long-horizon forecasting. A single shared model is trained across all carriers and sectors of an eNB, enabling efficient learning of joint traffic dynamics with low computational overhead. Forecast uncertainty is captured through quantile-based prediction intervals, providing confidence-aware estimates of future PRB availability. Evaluations on six months of commercial LTE network data from multiple U.S. locations demonstrate median MAE below 0.05 and hit probabilities above 0.80 for both one-day and seven-day recursive forecasts. These probabilistic predictions directly support spectrum-aware RAN functions such as dynamic carrier activation, congestion avoidance, and proactive spectrum sharing, making the proposed framework well-suited for dynamic spectrum access scenarios.

Problem

• residual PRB forecasting
• cross-carrier dependencies
• forecast uncertainty
• spectrum-aware control
• probabilistic prediction

Innovation

• Probabilistic Transformer
• Residual PRB Forecasting
• Cross-carrier Dependency
• Recursive Multivariate Time Series
• Quantile-based Uncertainty