Datasets:
metadata
license: unknown
task_categories:
- time-series-forecasting
tags:
- time-series
- forecasting
- long-horizon
size_categories:
- n<1K
pretty_name: etth2
etth2
Time series dataset: etth2
Dataset Description
- Homepage:
- Repository:
- Paper:
- License: Unknown
Dataset Summary
| Property | Value |
|---|---|
| Frequency | 1小时 |
| Validation samples | 1 |
| Train samples | 1 |
| Test samples | 1 |
Supported Tasks
- Time series forecasting
- Anomaly detection
- Classification (if applicable)
Languages
N/A (numerical data)
Dataset Structure
Data Instances
{
"item_id": "example_series_0",
"start": "2020-01-01T00:00:00",
"target": [1.0, 2.0, 3.0, ...],
"frequency": "1H",
"metadata": "{...}"
}
Data Fields
| Field | Type | Description |
|---|---|---|
| item_id | string | Unique identifier for the time series |
| start | string | ISO 8601 timestamp of the first observation |
| target | list[float] | Time series values |
| frequency | string | Pandas frequency string (e.g., '1H', '1D') |
| feat_dynamic_real | list[list[float]] | Time-varying covariates (optional) |
| feat_static_cat | list[int] | Static categorical features (optional) |
| metadata | string | JSON string with normalization params, etc. |
Data Splits
| Split | Examples |
|---|---|
| validation | 1 |
| train | 1 |
| test | 1 |
Dataset Creation
Source Data
Download Method: unknown
Preprocessing
- Data downloaded from original source
- Missing values filled using forward-fill method
- Standard normalization applied (mean=0, std=1)
- Split into train/validation/test sets (70/10/20)
- Converted to Parquet format for efficient streaming
Considerations for Using the Data
Social Impact
This dataset is intended for research purposes in time series forecasting.
Limitations
- Normalization parameters are computed on training data only
- Missing value handling may introduce artifacts
- Temporal alignment assumes regular intervals
Additional Information
Citation
@misc{unknown_dataset,
title = {Unknown Dataset},
url = {},
year = {2024},
}
Contributions
This dataset was processed and uploaded as part of the TS Arena benchmarking project.
Generated automatically by TS Arena streaming pipeline on 2026-01-03