README.md

---
license: mit
task_categories:
  - time-series-forecasting
task_ids:
  - univariate-time-series-forecasting
  - multivariate-time-series-forecasting
pretty_name: Beam-Level (5G) Time-Series Dataset
configs:
  - config_name: DLPRB
    data_files:
      - split: train
        path: data/train/DLPRB_train_0w-5w.csv
---

## Beam-Level (5G) Time-Series Dataset
This dataset presents a novel, multi-variate time series specifically designed for advancing research in spatio-temporal forecasting. Our primary goal is to facilitate the accurate prediction of traffic throughput volumes across communication networks, as visually depicted in Figure 1.

![An illustration of a base station (center), two cells (left and right), and four beams in each cell.](images/network.png)

The precise forecasting of network traffic volume is crucial for optimizing network flow management and efficient resource allocation. Consequently, this task holds significant practical and theoretical relevance for the scientific community in both networking and machine learning domains. The dataset aims to provide a valuable benchmark for researchers exploring state-of-the-art (SOTA) techniques in time series analysis.


## Dataset Split
This repository contains four datasets containing network performance metrics for 2,880 beams across 30 base stations. Each base station consists of 3 cells with 32 beams, with data recorded hourly. These datasets encompass a five-week period with data recorded at hourly intervals (as illustrated in Figure 2). These datasets are traffic_DLThpVol.csv, traffic_DLThpTime.csv, traffic_MR_number.csv, and traffic_DLPRB.csv. We remind the participants that the objective is to forecast future values of traffic volume (DLThpVol).

![Dataset splits: train set (first 5 weeks), and two target weeks (immediate 6th week and the 11th week).](images/dataset_split.png)

Each dataset corresponds to a specific network performance metric:

- traffic_DLThpVol.csv: represents throughput volume.
- traffic_DLThpTime.csv: represents throughput time.
- traffic_ DLPRB.csv: represents Physical Resource Block (PRB) utilization.
- traffic_MR_number.csv: represents user count.

### Citation

Please cite this paper if you intend to use this dataset for your research:
> L. Fechete et al., Goal-Oriented Time-Series Forecasting: Foundation Framework Design, arXiv:2504.17493 (2025).