Modeling configuration-performance relation in a mobile network: a data-driven approach

TL;DR

The paper addresses the challenge of predicting mobile network performance across a broad, multidimensional configuration space. It introduces the Configuration-Performance Modeling (CMPM) framework that fuses Configuration Management (CM) and Performance Management (PM) data from OSS to predict Downlink Throughput under varying configurations and environmental factors. CMPM, a 3-layer neural network, outperforms per-cell, fixed-configuration baselines and generalizes to unseen configurations, demonstrating the value of joint configuration-environment modeling for network optimization and digital twin applications. The work also discusses a scalable data-processing pipeline, data segmentation via change-point detection, and stability considerations under different network loads, highlighting practical implications for automated performance management and future enhancements such as hierarchical modeling and improved sampling strategies.

Abstract

Mobile network performance modeling typically assumes either a fixed cell's configuration or only considers a limited number of parameters. This prohibits the exploration of multidimensional, diverse configuration space for, e.g., optimization purposes. This paper presents a method for performance predictions based on a network cell's configuration and network conditions, which utilizes neural network architecture. We evaluate the idea by extensive experiments, with data from more than 50,000 5G cells. The assessment included a comparison of the proposed method against models developed for fixed configuration. Results show that combined configuration-performance modeling outperforms single-configuration models and allows for performance prediction of unknown configurations, i.e., it is not used for model training. A substantially lower mean absolute error was achieved (0.25 vs. 0.45 for fixed-configuration MLP-based models).

Figures (9)

• Figure 1: Distributions of Spearman's correlation coefficients among environmental metrics and downlink throughput.
• Figure 2: Distributions of downlink throughput per band recorded in live 5G network.
• Figure 3: Data processing pipeline for KPI modeling.
• Figure 4: Data segmentation with configuration data and change point detection in performance data. Shaded areas denote periods, where configuration is assumed to be stable.
• Figure 5: Illustration of evaluation approach. Whole block incorporates data from single cell-state. First 70% of data is used to train cell-state models (with only environmental data). Remaining 30% is used to evaluate both cell-state and CMPM models.