README.md

---
license: other
task_categories:
- time-series-forecasting
task_ids:
- univariate-time-series-forecasting
- multivariate-time-series-forecasting
annotations_creators:
- no-annotation
source_datasets:
- original
tags:
- forecasting
- benchmark
- fev
- arxiv:2509.26468
- tsfile
- modality:timeseries
- timeseries
- format:tsfile
size_categories:
- n<1K
pretty_name: boomlet (TsFile format)
configs:
- config_name: default
  data_files:
  - split: train
    path: "**/*.tsfile"
modality:
- timeseries
---

# boomlet (TsFile format)

This repository contains time-series forecasting data stored in [Apache TsFile](https://tsfile.apache.org/) format.

## Summary

- FEV subset: `boomlet`
- Unified source collection: [`autogluon/fev_datasets`](https://huggingface.co/datasets/autogluon/fev_datasets)
- Original source: https://huggingface.co/datasets/Datadog/BOOM
- Paper / citation: [[5]](https://arxiv.org/abs/2505.14766)
- Series: 1
- Modalities: Time-series
- TsFile rows (flattened observations): 10,139,745
- Frequencies: 1062, 1209, 1225, 1230, 1282, 1487, 1631, 1676, 1855, 1975, 2187, 285, 619, 772, 963
- TsFile files: 15
- Time precision: milliseconds (`INT64`).

Licensing and citation requirements follow the original source. This repository does not claim ownership of the original data.

## Dataset Statistics

| Frequency | Series | Median series length | TsFile rows (observations) | Dynamic columns | Static columns | Data files |
|---|---:|---:|---:|---:|---:|---|
| 1062 | 1 | 16,384 | 344,064 | 21 | 6 | `1062/1062.tsfile` |
| 1209 | 1 | 16,384 | 868,352 | 53 | 6 | `1209/1209.tsfile` |
| 1225 | 1 | 16,384 | 802,816 | 49 | 6 | `1225/1225.tsfile` |
| 1230 | 1 | 16,384 | 376,832 | 23 | 6 | `1230/1230.tsfile` |
| 1282 | 1 | 16,384 | 573,440 | 35 | 6 | `1282/1282.tsfile` |
| 1487 | 1 | 16,384 | 884,736 | 54 | 6 | `1487/1487.tsfile` |
| 1631 | 1 | 10,463 | 418,520 | 40 | 6 | `1631/1631.tsfile` |
| 1676 | 1 | 10,463 | 1,046,300 | 100 | 6 | `1676/1676.tsfile` |
| 1855 | 1 | 5,231 | 272,012 | 52 | 6 | `1855/1855.tsfile` |
| 1975 | 1 | 5,231 | 392,325 | 75 | 6 | `1975/1975.tsfile` |
| 2187 | 1 | 5,231 | 523,100 | 100 | 6 | `2187/2187.tsfile` |
| 285 | 1 | 16,384 | 1,228,800 | 75 | 6 | `285/285.tsfile` |
| 619 | 1 | 16,384 | 851,968 | 52 | 6 | `619/619.tsfile` |
| 772 | 1 | 16,384 | 1,097,728 | 67 | 6 | `772/772.tsfile` |
| 963 | 1 | 16,384 | 458,752 | 28 | 6 | `963/963.tsfile` |

## Files

The Hugging Face dataset card YAML points `configs.data_files` to all `*.tsfile` files in this repository.

- `1062/1062.tsfile`
- `1209/1209.tsfile`
- `1225/1225.tsfile`
- `1230/1230.tsfile`
- `1282/1282.tsfile`
- `1487/1487.tsfile`
- `1631/1631.tsfile`
- `1676/1676.tsfile`
- `1855/1855.tsfile`
- `1975/1975.tsfile`
- `2187/2187.tsfile`
- `285/285.tsfile`
- `619/619.tsfile`
- `772/772.tsfile`
- `963/963.tsfile`

## TsFile Storage Model

- Each original series (`id`) is stored as one TsFile device.
- Static covariate columns are stored as TAG columns: `type, Application_Usage, Infrastructure, Database, Networking, Security`.
- Time-varying targets and dynamic covariates are stored as FIELD measurements.
- Source `timestamp` values are mapped to the TsFile `Time` column as millisecond timestamps.
- Table name(s): boomlet_1062, boomlet_1209, boomlet_1225, boomlet_1230, boomlet_1282, boomlet_1487, boomlet_1631, boomlet_1676, boomlet_1855, boomlet_1975, boomlet_2187, boomlet_285, boomlet_619, boomlet_772, boomlet_963.

### Column Schema

| Column | Role | TsFile type |
|---|---|---|
| `Time` | Time column | INT64 |
| `id` | TAG (device dimension) | STRING |
| `type` | TAG (device dimension) | STRING |
| `Application_Usage` | TAG (device dimension) | DOUBLE |
| `Infrastructure` | TAG (device dimension) | DOUBLE |
| `Database` | TAG (device dimension) | DOUBLE |
| `Networking` | TAG (device dimension) | DOUBLE |
| `Security` | TAG (device dimension) | DOUBLE |
| `target_0` | FIELD (measurement) | FLOAT |
| `target_1` | FIELD (measurement) | FLOAT |
| `target_2` | FIELD (measurement) | FLOAT |
| `target_3` | FIELD (measurement) | FLOAT |
| `target_4` | FIELD (measurement) | FLOAT |
| `target_5` | FIELD (measurement) | FLOAT |
| `target_6` | FIELD (measurement) | FLOAT |
| `target_7` | FIELD (measurement) | FLOAT |
| `target_8` | FIELD (measurement) | FLOAT |
| `target_9` | FIELD (measurement) | FLOAT |
| `target_10` | FIELD (measurement) | FLOAT |
| `target_11` | FIELD (measurement) | FLOAT |
| `target_12` | FIELD (measurement) | FLOAT |
| `target_13` | FIELD (measurement) | FLOAT |
| `target_14` | FIELD (measurement) | FLOAT |
| `target_15` | FIELD (measurement) | FLOAT |
| `target_16` | FIELD (measurement) | FLOAT |
| `target_17` | FIELD (measurement) | FLOAT |
| `target_18` | FIELD (measurement) | FLOAT |
| `target_19` | FIELD (measurement) | FLOAT |
| `target_20` | FIELD (measurement) | FLOAT |

> Note: 15 original `id` values contained invalid identifier characters and were normalized to valid device names, for example 1062→_1062, 1209→_1209, 1225→_1225.

## Conversion Notes

- The source FEV format stores each time series as one nested row containing `id`, `timestamp[]`, and target or covariate arrays.
- The TsFile conversion flattens those nested arrays into long rows. Therefore, the `TsFile rows` values above correspond to the number of timestamped observations after flattening.
- TAG columns identify the device and static metadata. FIELD columns contain values that change over time.
- Large logical tables may be split into multiple `.tsfile` shards such as `<name>_1.tsfile`, `<name>_2.tsfile`, and so on. Shards listed for the same frequency belong to the same logical table.

## Reading Example

```python
from tsfile import TsFileReader

reader = TsFileReader("1062/1062.tsfile")
schemas = reader.get_all_table_schemas()
# Table name(s): boomlet_1062, boomlet_1209, boomlet_1225, boomlet_1230, boomlet_1282, boomlet_1487, boomlet_1631, boomlet_1676, boomlet_1855, boomlet_1975, boomlet_2187, boomlet_285, boomlet_619, boomlet_772, boomlet_963
```