JamesBedichek/stock-prediction-data

---
license: mit
task_categories:
- tabular-regression
- tabular-classification
tags:
- finance
- stocks
- time-series
- trading
size_categories:
- 10B<n<100B
---

# Stock Prediction Dataset

This dataset contains preprocessed stock market data for training stock return prediction models.

## Contents

- `data/all_complete_dataset.h5` - Main features dataset (HDF5)
- ~4000 stocks
- 20+ years of daily data
- Technical indicators, fundamentals, news sentiment

- `data/actual_prices.h5` - Price data for backtesting
- Split-adjusted close prices

- `checkpoints/walk_forward/` - Pre-trained model checkpoints
- 6-fold walk-forward validated models
- Transformer-based architecture

## Usage

```python
from huggingface_hub import hf_hub_download

# Download the main dataset
hf_hub_download(
repo_id="JamesBedichek/stock-prediction-data",
filename="data/all_complete_dataset.h5",
repo_type="dataset",
local_dir="."
)
```

Or use the provided download script:
```bash
python scripts/download_data.py --repo-id JamesBedichek/stock-prediction-data
```

## License

MIT