---
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: us_consumption (TsFile format)
configs:
- config_name: default
  data_files:
  - split: train
    path: "**/*.tsfile"
modality:
- timeseries
---

# us_consumption (TsFile format)

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

## Summary

- FEV subset: `us_consumption`
- Unified source collection: [`autogluon/fev_datasets`](https://huggingface.co/datasets/autogluon/fev_datasets)
- Original source: https://apps.bea.gov/iTable/?reqid=19&step=3&isuri=1&nipa_table_list=2017&categories=underlying
- Paper / citation: [[23]](https://doi.org/10.1016/j.ijforecast.2016.04.005)
- Series: 31
- Modalities: Time-series
- TsFile rows (flattened observations): 34,658
- Frequencies: 1M, 1Q, 1Y
- TsFile files: 3
- 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 |
|---|---:|---:|---:|---:|---:|---|
| 1M | 31 | 792 | 24,552 | 1 | 0 | `1M/1M.tsfile` |
| 1Q | 31 | 262 | 8,122 | 1 | 0 | `1Q/1Q.tsfile` |
| 1Y | 31 | 64 | 1,984 | 1 | 0 | `1Y/1Y.tsfile` |

## Files

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

- `1M/1M.tsfile`
- `1Q/1Q.tsfile`
- `1Y/1Y.tsfile`

## TsFile Storage Model

- Each original series (`id`) is stored as one TsFile device.
- 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): us_consumption_1M, us_consumption_1Q, us_consumption_1Y.

### Column Schema

| Column | Role | TsFile type |
|---|---|---|
| `Time` | Time column | INT64 |
| `id` | TAG (device dimension) | STRING |
| `target` | FIELD (measurement) | FLOAT |

> Note: 3 original `id` values contained invalid identifier characters and were normalized to valid device names, for example food_and_beverages_purchased_for_off-premises_consumption→food_and_beverages_purchased_for_off_premises_consumption, food_and_beverages_purchased_for_off-premises_consumption→food_and_beverages_purchased_for_off_premises_consumption, food_and_beverages_purchased_for_off-premises_consumption→food_and_beverages_purchased_for_off_premises_consumption.

## 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("1M/1M.tsfile")
schemas = reader.get_all_table_schemas()
# Table name(s): us_consumption_1M, us_consumption_1Q, us_consumption_1Y
```