klines-gateio / README.md
rogerdehe's picture
add dataset card
adea16c verified
|
Raw
History Blame Contribute Delete
2.33 kB
---
license: mit
task_categories:
- time-series-forecasting
language:
- en
tags:
- crypto
- cryptocurrency
- finance
- ohlcv
- klines
- gateio
- trading
pretty_name: Gate.io USDT OHLCV Klines
size_categories:
- 1B<n<10B
---
# Gate.io USDT OHLCV Klines
Historical OHLCV candlestick (kline) data for **Gate.io** USDT-quoted markets,
stored as zstd-compressed Parquet.
## Coverage
| | |
|---|---|
| **Exchange** | Gate.io |
| **Markets** | `spot` (primary) · `futures` (USDT perpetuals — limited history, see Notes) |
| **Timeframes** | 5m · 15m · 1h · 4h · 1d (all pairs) — plus 1m for a ~1.3k-pair subset |
| **Pairs** | ~3,300 spot USDT pairs |
| **History** | spot back to **2013** for the oldest listings |
| **As of** | static snapshot (~2025-05); periodic refresh planned |
| **Source** | seeded from the DigiTuccar `HistoricalDataForTradeBacktest` snapshot |
## File layout
```
{market}/{timeframe}/{PAIR}.parquet
```
Examples: `spot/1d/BTC_USDT.parquet`, `spot/5m/ETH_USDT.parquet`
Each Parquet holds **one pair × one timeframe**. Columns:
| column | type | description |
|---|---|---|
| `date` | datetime (UTC) | candle open time |
| `open` | float | open price |
| `high` | float | high price |
| `low` | float | low price |
| `close` | float | close price |
| `volume` | float | base-asset volume |
## Usage
Single file:
```python
from huggingface_hub import hf_hub_download
import pandas as pd
path = hf_hub_download("rogerdehe/klines-gateio",
"spot/1d/BTC_USDT.parquet", repo_type="dataset")
df = pd.read_parquet(path)
```
A subset (or all) of the dataset:
```python
from huggingface_hub import snapshot_download
snapshot_download("rogerdehe/klines-gateio", repo_type="dataset",
allow_patterns=["spot/1d/*"]) # e.g. just daily spot
```
## Notes
- **Futures history is limited**: Gate.io's API only served ~180 days back at
snapshot time, so `futures/` is thin relative to `spot/`. Use spot for deep history.
- `1m` is available only for a subset of pairs (~1.3k); other timeframes cover all pairs.
- The HF **dataset viewer is disabled** (custom per-pair layout); load files directly as shown above.
- Timestamps are UTC. Data is provided **as-is** for research purposes, with no
warranty of accuracy or completeness.