Predicting Parameter Change's Effect on Cellular Network Time Series

TL;DR

This work tackles predicting how cell parameter adjustments affect cellular-time-series metrics, enabling proactive parameter tuning. It introduces ParaSeer, which splits the task into adjustment-free forecasting using a Graphical Transformer and a multiplier-based adjustment derived from a homogeneous single-cell model, formalized as a multiplier α(β|θ). A Graphical Transformer framework, enhanced by the Homogeneous Single-Cell Parameter Effect Predictor, propagates parameter-change effects through a causal graph and leverages domain knowledge to reduce data needs. The authors derive an interpretable multiplier based on transmission power and CIO and demonstrate strong empirical gains over baselines on two real-world datasets, with robust generalization to unseen cells and new regions. This approach offers a practical, explainable pathway for anticipatory network optimization in evolving cellular systems.

Abstract

The cellular network provides convenient network access for ever-growing mobile phones. During the continuous optimization, operators can adjust cell parameters to enhance the Quality of Service (QoS) flexibly. A precise prediction of the parameter change's effect can help operators make proper parameter adjustments. This work focuses on predicting cell status (like the workload and QoS) after adjusting the cell parameters. The prediction will be conducted before an adjustment is actually applied to provide an early inspection. As it can be hard for available parameter adjustments with a limited number to cover all the parameter and user behavior combinations, we propose ParaSeer fusing domain knowledge on parameter adjustments into data-driven time series forecasting. ParaSeer organizes several pre-trained Transformers for adjustment-free time series forecasting, utilizing plenty of adjustment-free data. On the other hand, ParaSeer models the effect of adjusting the transmission power and cell individual offset (CIO) as a multiplier for the workload. We derive a formula to calculate the multiplier from the underlying mechanism of those two parameters, helping ParaSeer eliminate the thirst for data with parameter adjustments. We compare ParaSeer with baselines on two real-world datasets, where ParaSeer outperforms the best baseline by more than 25.8% in terms of RMSE. The extensive experiments further illustrate the contributions of ParaSeer's components.

Figures (11)

• Figure 1: Time series of the maximum number of RRC connections around an adjustment from a real-world cell. The dashed vertical line indicates the first data point with the new transmission power.
• Figure 2: Graphical model for a single cell. Dashed arrows from time encode the periodicity of the related variables.
• Figure 3: Two omnidirectional cells. As $A$'s transmission power changes from $\tau$ dBm to $\tau + \delta$ dBm and its CIO changes from $o$ dB to $o + \delta_{o}$ dB, a UE connected to $A$ will observe a change in the offset RSRP from $P_{R}$ to $P_{R}^{\prime} = \beta P_{R}$, where $\beta = 10^{0.1 (\delta + \delta_{o})}$. With $\beta < 1$, $A$'s boundary shrinks to the dotted circle.
• Figure 4: The serving region of the concerned gNB with different ratios between its offset transmission power and that of its neighbors. $\theta$ is the angular interval between two neighbors.
• Figure 5: Architecture of ParaSeer. ParaSeer aims to fill the empty dotted boxes in the left part with its outputs based on the available data. The Graphical Transformer module is responsible for time series forecasting. The HSC-PEP (Homogeneous Single-Cell Parameter Effect Predictor) module converts the predicted adjustment-free workload into the one with adjustment, which is further sent to the other Transformers.

Theorems & Definitions (2)

• Theorem 1
• proof : Proof of Theorem \ref{['thm:track-necessity']}