Our team at FIIT STU, together with experts from Ericsson Slovakia and with data provided by CETIN Networks, has developed an open dataset from a live commercial mobile network in Slovakia. Accepted for publication in Scientific Data, the work combines internal performance counters from 2G, 4G and 5G networks with real energy consumption measurements, creating a valuable resource for data-driven network optimisation.
Addressing the Lack of Real Network Data
Developing more efficient mobile networks requires understanding how traffic demand, radio conditions and infrastructure energy consumption interact. However, internal network measurements are rarely available outside operators due to confidentiality and operational constraints.
Researchers therefore often rely on simulations or measurements collected by mobile devices. While useful, these resources offer limited visibility into the infrastructure supporting the connection. Our dataset addresses this gap by making anonymised internal network telemetry openly available for research.
What Makes This Dataset Different?
The novelty lies in combining three mobile network generations, multiple frequency bands, 15-minute measurement intervals and actual infrastructure energy consumption within a single openly available resource.
Unlike datasets based on drive tests or user-device observations, it provides performance counters collected directly from operational base stations. These include radio resource utilisation, connected and active user counts, transferred data volumes, channel quality indicators and MIMO transmission ranks, alongside energy measurements.
According to the authors’ review of existing resources, this is the first publicly available dataset to offer this particular combination of raw multi-technology network counters and real energy measurements from a live commercial network.
The anonymisation preserves relationships between base stations, baseband units, radio units and sectors, allowing researchers to investigate interactions between technologies and frequency bands sharing the same infrastructure.
More Than 34 Million Measurement Records
The released resource contains 34,117,453 records across three dataset partitions. The largest partition covers 264 base stations over 84 days, contributing more than 30 million records.
The team checked the data for completeness, format consistency, duplicate records and technical plausibility. Missing measurements and operational outliers were deliberately retained to preserve real network behaviour. Approximately 17% of records contain at least one missing performance indicator, providing opportunities to evaluate methods that handle incomplete operational data.
Available as CSV files on Zenodo under a CC BY 4.0 licence, the dataset supports research into:
- Traffic and energy demand forecasting.
- Energy-aware network management and capacity planning.
- Anomaly detection and network diagnostics.
- Explainable machine learning and cellular network digital twins.
The main outcome is an openly accessible, validated foundation for developing and evaluating these approaches. Energy measurements describe shared infrastructure components, so analyses must account for the fact that consumption cannot be directly attributed to individual network cells.
Why Publication in Scientific Data Matters
Scientific Data is a peer-reviewed, open-access journal within Nature Portfolio, specialising in scientifically valuable research datasets. Its review process evaluates both the manuscript and the associated data, focusing on technical quality, appropriate collection and processing methods, documentation, validation and suitability for reuse.
Acceptance recognises the dataset as a documented research contribution that other teams can examine and build upon. By opening access to measurements previously confined to commercial infrastructure, this work supports reproducible telecommunications research and further investigation of energy-efficient mobile networks.
Authors: Peter Lehoczký, Matúš Turcsány, Laura Krajčovičová, Filip Zatroch, Marcel Kajan and Marek Galinski.
Read the paper: Performance Management Counters from Live 5G, 4G and 2G Radio Access Network
Access the dataset: Zenodo