TrungKiencding/VT_Telco_TS_Dataset

VT Telco Time Series Dataset

Time series dataset extracted from IPMS network telemetry for forecasting and monitoring tasks in telecom infrastructure.

The dataset contains one row per time series in Arrow IPC format and is intended for use with GluonTS/Chronos-style modeling workflows.

Data format and usage

Each dataset row corresponds to one univariate time series with the following schema:

• item_id (string): interface name.
• series_id (string): series identifier parsed from the raw source key.
• start (timestamp[us]): start timestamp of the series.
• target (Sequence[float32]): observed values for that series.

Current release summary:

• Number of series (rows): 1118
• Fixed history length per series: 2016 points
• File size: 8,667,418 bytes (~8.27 MB)
• Earliest start timestamp: 2025-05-09 23:55:00
• Latest start timestamp: 2025-05-10 23:40:00

Load with PyArrow

import pyarrow as pa
import pyarrow.ipc as ipc

with pa.memory_map("ipms.arrow", "r") as source:
    table = ipc.open_file(source).read_all()

print(table.schema)
print(table.slice(0, 1).to_pylist()[0])

Convert to pandas

import pyarrow as pa
import pyarrow.ipc as ipc

with pa.memory_map("ipms.arrow", "r") as source:
    table = ipc.open_file(source).read_all()

df = table.to_pandas()
print(df.head())

Sample entry

{
  "item_id": "GigabitEthernet0/6",
  "series_id": "104741782",
  "start": "2025-05-09 23:55:00",
  "target": [132494112.0, 107854768.0, 85698952.0, ...]
}

Data preparation

The source CSV is converted to Arrow using scripts/ipms_to_gluonts_arrow.py with these key processing steps:

• Keep only rows where cc_anomaly_result is in {0, false} (case-insensitive).
• Parse interface field into series_id, item_id, and fallback timestamp.
• Use up to 2016 points per series (configured by --max-points at generation time).
• Write Arrow IPC with compression (lz4 by default).

Changelog

• v1.0.0 (2026-04-03): Initial public release of ipms.arrow for VT Telco forecasting.

License

This dataset is currently released as other license.
Please update this section with the exact license terms before production or external redistribution.

Citation

If you use this dataset, please cite the dataset repository:

@dataset{trungkien_vt_telco_ts_2026,
  author = {TrungKiencding},
  title = {VT Telco Time Series Dataset},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/TrungKiencding/VT_Telco_TS_Dataset}
}