TelcoAgent: A Scalable 5G Multi-KPM Forecasting With 3GPP-Grounded Explainability

Researchers introduce TelcoAgent, a foundation model-based framework that forecasts multiple Key Performance Measurements (KPMs) across 5G networks with high accuracy and explainability. The system leverages 3GPP knowledge graphs and time-series foundation models to enable zero-shot forecasting across diverse network cells without site-specific retraining, validated on real-world city-scale 5G data.

TelcoAgent addresses a critical operational challenge in 5G network management: predicting performance degradations before they impact users. Traditional machine learning approaches to KPM forecasting suffer from poor scalability and lack transparency into their predictions, limiting adoption by telecom operators who require explainable diagnostics to justify network interventions. This research bridges that gap by combining three complementary capabilities: automated knowledge extraction from 3GPP standards into structured knowledge graphs, zero-shot forecasting using time-series foundation models, and rule-based reasoning that ties predictions back to domain-specific factors.

The achievement demonstrates meaningful progress in applying foundation models beyond their original domains. Rather than treating telecom as just another time-series forecasting problem, the researchers engineered domain grounding through 3GPP standards, enabling the system to provide actionable diagnostics rather than black-box predictions. Real-world validation across 200 cells tracking 7 different KPMs over three months shows the approach handles complexity that would typically require extensive site-specific tuning.

For telecom operators, this framework could reduce operational overhead by automating routine forecasting and diagnostics while maintaining regulatory compliance through explainability. The zero-shot capability is particularly valuable—deploying to new network cells typically requires months of historical data collection and model retraining. By eliminating this bottleneck, operators can accelerate network optimization and proactive maintenance across their entire infrastructure.

The technical approach establishes a template for applying foundation models to regulated industries where explainability requirements have traditionally favored simpler models. Future work likely focuses on extending the reasoning pipeline to recommend specific remediation actions and scaling to international operators with different 3GPP implementations.

• →Foundation models can achieve zero-shot forecasting across diverse 5G network cells without site-specific retraining, reducing deployment complexity.
• →3GPP knowledge graphs extracted from standards documents enable domain-grounded explanations that satisfy telecom operators' need for interpretable diagnostics.
• →Real-world validation across 200 cells demonstrates practical viability for forecasting seven different KPM types with high accuracy.
• →The framework combines time-series foundation models with automated reasoning to bridge the gap between accuracy and explainability in critical infrastructure.
• →Elimination of site-specific training requirements could accelerate proactive network management and reduce operational burden for telecom providers.