Datasets:
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
task_categories:
|
| 4 |
+
- time-series-forecasting
|
| 5 |
+
- regression
|
| 6 |
+
tags:
|
| 7 |
+
- financial
|
| 8 |
+
- prediction-market
|
| 9 |
+
- polymarket
|
| 10 |
+
- time-series
|
| 11 |
+
size_categories:
|
| 12 |
+
- 10M<n<100M
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Polymarket时间序列预测数据集
|
| 16 |
+
|
| 17 |
+
## 数据集描述
|
| 18 |
+
|
| 19 |
+
这是一个从Polymarket提取的时间序列预测数据集,用于训练WaveNet模型预测未来市场走势。
|
| 20 |
+
|
| 21 |
+
### 数据集统计
|
| 22 |
+
|
| 23 |
+
- **总样本数**: 6,780,504
|
| 24 |
+
- **特征维度**: 1,500 (100个时间段 × 15个因子)
|
| 25 |
+
- **标签维度**: 10 (预测未来10个时间段)
|
| 26 |
+
- **文件大小**: ~12.36 GB
|
| 27 |
+
- **格式**: Parquet
|
| 28 |
+
|
| 29 |
+
### 特征描述
|
| 30 |
+
|
| 31 |
+
每个样本包含:
|
| 32 |
+
- **特征**: 100个历史时间段的15个因子特征 (1,500维)
|
| 33 |
+
- **标签**: 未来10个时间段的response_zscore (10维)
|
| 34 |
+
- **市场ID**: 标识数据来源的市场
|
| 35 |
+
|
| 36 |
+
### 15个因子
|
| 37 |
+
|
| 38 |
+
1. price_std_zscore - 价格标准差
|
| 39 |
+
2. price_trend_zscore - 价格趋势
|
| 40 |
+
3. price_skew_zscore - 价格偏度
|
| 41 |
+
4. price_range_zscore - 价格范围
|
| 42 |
+
5. vol_sum_abs_zscore - 成交量总和
|
| 43 |
+
6. vol_std_zscore - 成交量标准差
|
| 44 |
+
7. vol_flow_zscore - 成交量流
|
| 45 |
+
8. vwap_zscore - 成交量加权平均价
|
| 46 |
+
9. weighted_volatility_zscore - 加权波动率
|
| 47 |
+
10. vol_price_corr_zscore - 量价相关性
|
| 48 |
+
11. energy_zscore - 能量
|
| 49 |
+
12. flow_ratio_zscore - 流动比率
|
| 50 |
+
13. imbalance_intensity_zscore - 不平衡强度
|
| 51 |
+
14. vwap_diff_prev_zscore - VWAP差分
|
| 52 |
+
15. vwap_slope_ma5_zscore - VWAP斜率
|
| 53 |
+
|
| 54 |
+
### 数据分割
|
| 55 |
+
|
| 56 |
+
- **训练集**: 80%
|
| 57 |
+
- **验证集**: 10%
|
| 58 |
+
- **测试集**: 10%
|
| 59 |
+
|
| 60 |
+
## 使用方法
|
| 61 |
+
|
| 62 |
+
### 使用Hugging Face Datasets
|
| 63 |
+
|
| 64 |
+
```python
|
| 65 |
+
from datasets import load_dataset
|
| 66 |
+
|
| 67 |
+
dataset = load_dataset("chaoleiyv/marketdata", split="train")
|
| 68 |
+
```
|
| 69 |
+
|
| 70 |
+
### 直接使用Parquet文件
|
| 71 |
+
|
| 72 |
+
```python
|
| 73 |
+
import pandas as pd
|
| 74 |
+
|
| 75 |
+
df = pd.read_parquet("ml_dataset.parquet")
|
| 76 |
+
```
|
| 77 |
+
|
| 78 |
+
## 数据集结构
|
| 79 |
+
|
| 80 |
+
```
|
| 81 |
+
ml_dataset/
|
| 82 |
+
├── ml_dataset.parquet # 完整数据集(Parquet格式)
|
| 83 |
+
└── README.md # 数据集说明
|
| 84 |
+
```
|
| 85 |
+
|
| 86 |
+
## 引用
|
| 87 |
+
|
| 88 |
+
如果使用本数据集,请引用:
|
| 89 |
+
|
| 90 |
+
```bibtex
|
| 91 |
+
@dataset{polymarket_ml_dataset,
|
| 92 |
+
title={Polymarket时间序列预测数据集},
|
| 93 |
+
author={chaoleiyv},
|
| 94 |
+
year={2025},
|
| 95 |
+
url={https://huggingface.co/datasets/chaoleiyv/marketdata}
|
| 96 |
+
}
|
| 97 |
+
```
|
| 98 |
+
|
| 99 |
+
## 许可证
|
| 100 |
+
|
| 101 |
+
MIT License
|