Multivariate, Multi-step, and Spatiotemporal Traffic Prediction for NextG Network Slicing under SLA Constraints

TL;DR

This paper introduces a multivariate, multi-step spatiotemporal traffic prediction method for NextG networks that ensures SLA compliance, comparing single-cell, multi-cell, and slice-based approaches with significant performance improvements.

Contribution

The study proposes a novel SLA-aware loss function and evaluates different training architectures, demonstrating the superiority of single-cell and multi-slice models for traffic prediction.

Findings

Single-cell training outperforms multi-cell with 11.4% and 38.1% test loss improvements.

Slice-based prediction significantly reduces test loss by up to 55.6%.

Multi-slice models provide more accurate traffic forecasts than single-slice models.

Abstract

This study presents a spatiotemporal traffic prediction approach for NextG mobile networks, ensuring the service-level agreements (SLAs) of each network slice. Our approach is multivariate, multi-step, and spatiotemporal. Leveraging 20 radio access network (RAN) features, peak traffic hour data, and mobility-based clustering, we propose a parametric SLA-based loss function to guarantee an SLA violation rate. We focus on single-cell, multi-cell, and slice-based prediction approaches and present a detailed comparative analysis of their performances, strengths, and limitations. First, we address the application of single-cell and multi-cell training architectures. While single-cell training offers individual cell-level prediction, multi-cell training involves training a model using traffic from multiple cells from the same or different base stations. We show that the single-cell approach outperforms the multi-cell approach and results in test loss improvements of 11.4% and 38.1% compared to baseline SLA-based and MAE-based models, respectively. Next, we explore slice-based traffic prediction. We present single-slice and multi-slice methods for slice-based downlink traffic volume prediction, arguing that multi-slice prediction offers a more accurate forecast. The slice-based model we introduce offers substantial test loss improvements of 28.2%, 36.4%, and 55.6% compared to our cell-based model, the baseline SLA-based model, and the baseline MAE-based model, respectively.