Hugging Face's logo

Datasets:
yinelon
/
crypto_lob_10m

Modalities:
Tabular
Text
Formats:
json
Size:
< 1K
ArXiv:
Libraries:
Datasets
pandas
Dataset card Data Studio Files
xet
Community

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

Crypto LOB Dataset for DeepLOB

High-frequency cryptocurrency Limit Order Book (LOB) dataset for training DeepLOB and other deep learning models.

📊 Dataset Overview

  • Symbols: 30 cryptocurrency perpetual futures
  • Time Range: August 1, 2025 - November 1, 2025 (3 months)
  • Frequency: 1-second snapshots
  • Total Samples: 232.8 million records
  • Size: ~36 GB
  • Format: NumPy .npy binary files
  • Data Source: Binance Futures via Tardis.dev API

📁 Data Format

Each .npy file has shape (N, 41):

Column Description
0 Unix timestamp (milliseconds)
1-40 Order book features (40 dimensions)
41 10-second forward return (label)

Order Book Features:

  • Columns 1-10: Ask prices [Level 1-10]
  • Columns 11-20: Ask volumes [Level 1-10]
  • Columns 21-30: Bid prices [Level 1-10]
  • Columns 31-40: Bid volumes [Level 1-10]

Label: return_10s = (mid_price_{t+10s} - mid_price_t) / mid_price_t

🪙 Included Symbols (30)

Major: BTC, ETH, BNB, SOL, XRP, DOGE

DeFi & Layer 1/2: AVAX, DOT, LINK, NEAR, UNI, FIL, LTC, ARB, TAO, SUI, WLD

Meme: PEPE (1000PEPE), WIF, TRUMP, PENGU, HYPE, PNUT (via PUMP), FARTCOIN

Others: BCH, AAVE, ZEC, ADA, ENA, MYX

🔧 Usage 

import numpy as np

# Load data
data = np.load('BTCUSDT_20250801_20251101.npy')

# Extract features and labels
features = data[:, 1:41]  # 40 features
returns = data[:, 41]     # 10-second returns

# Create sequences for DeepLOB
seq_len = 100
X = np.array([features[i:i+seq_len] for i in range(len(features) - seq_len)])
y = returns[seq_len:]

📚 Model Architecture

DeepLOB Input: (batch_size, 1, sequence_length, 40)
Architecture: CNN → Inception → LSTM
Output: Regression (returns) or Classification (up/down/stationary)

Reference: DeepLOB Paper

📈 Statistics

  • Per Symbol: ~7.8M samples, ~1.2 GB
  • Total: 232.8M samples, 36 GB
  • Feature Dimension: 40
  • Label Type: Regression (10-second returns)

🔗 Links

Downloads last month
6