📶 Beam-Level (5G) Time-Series Dataset

📚 Citation

This dataset is released alongside the following paper:

Fechete, L., et al. “Goal-Oriented Time-Series Forecasting: Foundation Framework Design.” Proceedings of the AAAI Conference on Artificial Intelligence, 2026, Singapore.

If you use this dataset, please cite the above work.

---

This dataset introduces a novel multivariate time series specifically curated to support research in enabling accurate prediction of KPIs across communication networks, as illustrated below:

 

Precise forecasting of network traffic is critical for optimizing network management and enhancing resource allocation efficiency. This task is of both practical and theoretical importance to researchers in networking and machine learning, offering a strong benchmark for state-of-the-art (SOTA) time series models.

---

📂 Dataset Overview

The dataset comprises:

• 2,880 Beams across 30 Base Stations (3 Cells per Station, 32 Beams per Cell).
• Duration: 5 weeks + 2 target weeks, totaling up to 840 training hours and 1176 total hours per beam.

---

📁 Available CSV Files

🏋️‍♂️ Training Set (Weeks 0–5)

🎯 Forecast Targets

📆 6th Week (Week 5–6)

📆 11th Week (Week 10–11)

---

🧪 Dataset Splits

 

The dataset is split into a Training Set (first 5 weeks) and Forecast Targets for Week 6 (immediate future) and Week 11 (long-term future).

---

📄 Data Format

Each CSV file contains a Time column and multiple beam columns (e.g., 0_0_0 to 29_2_31). The Time column ranges from 0–839 for training (weeks 1–6), 0–167 for week 6, and 168–335 for week 11. Each beam column uniquely identifies one of the 2,880 beams across 30 base stations.

---

🔗 Code Repository

The official codebase for working with this dataset is available here: 👉 https://github.com/netop-team/gotsf