Characterizing 5G User Throughput via Uncertainty Modeling and Crowdsourced Measurements

TL;DR

This work tackles the challenge of characterizing application-layer throughput in 5G by leveraging large-scale crowdsourced measurements to enable end-to-end QoS visibility. It adopts an uncertainty-aware framework using NGBoost to produce calibrated predictive intervals alongside point estimates, and compares against a 4G-based baseline (XGBoost). The study demonstrates improved throughput prediction, first benchmarks for 5G NSA and SA, and shows that bottlenecks shift from the RAN to transport and service layers as networks evolve, with E2E metrics becoming more influential. These findings have practical implications for QoS-aware network management and highlight the value of enhanced observability and uncertainty quantification in data-limited mobile networks.

Abstract

Characterizing application-layer user throughput in next-generation networks is increasingly challenging as the higher capacity of the 5G Radio Access Network (RAN) shifts connectivity bottlenecks towards deeper parts of the network. Traditional methods, such as drive tests and operator equipment counters, are costly, limited, or fail to capture end-to-end (E2E) Quality of Service (QoS) and its variability. In this work, we leverage large-scale crowdsourced measurements-including E2E, radio, contextual and network deployment features collected by the user equipment (UE)-to propose an uncertainty-aware and explainable approach for downlink user throughput estimation. We first validate prior 4G methods, improving R^2 by 8.7%, and then extend them to 5G NSA and 5G SA, providing the first benchmarks for 5G crowdsourced datasets. To address the variability of throughput, we apply NGBoost, a model that outputs both point estimates and calibrated confidence intervals, representing its first use in the field of computer communications. Finally, we use the proposed model to analyze the evolution from 4G to 5G SA, and show that throughput bottlenecks move from the RAN to transport and service layers, as seen by E2E metrics gaining importance over radio-related features.

Figures (4)

• Figure 1: NGBoost 95% confidence intervals for 50 random 5G SA samples. Dots show XGBoost point estimates (orange) and NGBoost means (blue); intervals are green when containing the true value (black cross) and red otherwise.
• Figure 2: Reliability diagram with the calibration curves for the NGBoost model for the different radio access technologies' testing sets, where $\alpha$ is the miscoverage level. The C-AUC is indicated in the legend.
• Figure 3: Average feature importance plot for the XGBoost and NGBoost models for 4G, 5G NSA and 5G SA. Feature categories are color-coded: radio (blue), E2E (red), contextual (green), and deployment (gray).
• Figure 4: SHAP vs feature value plots of XGBoost and NGBoost for the three most important inputs in 5G SA networks.