README.md

---
license: apache-2.0
task_categories:
  - time-series-forecasting
  - text-classification
tags:
  - finance
  - stock-trading
  - llm
  - backtesting
pretty_name: FINSABER Data
size_categories:
  - 10B<n<100B
---

# FINSABER Data

Aggregated datasets for the [FINSABER](https://github.com/waylonli/FINSABER) backtesting framework (KDD 2026).

## Files

| File | Description | Size |
| :--- | :--- | :--- |
| `data/finmem_data/stock_data_sp500_2000_2024.pkl` | S&P500 full aggregated data (Price + News + Filings) | ~11 GB |
| `data/finmem_data/stock_data_cherrypick_2000_2024.pkl` | Selected symbols (TSLA, AMZN, MSFT, NFLX, COIN) | ~53 MB |
| `data/price/all_sp500_prices_2000_2024_delisted_include.csv` | CSV price-only data for S&P500 (including delisted) | ~253 MB |

## Aggregated Data Structure (`.pkl`)

Each `.pkl` file is a Python dictionary keyed by `datetime.date`:

```python
{
    datetime.date(2024, 1, 2): {
        "price": {
            "AAPL": {
                "open": 187.15,
                "high": 188.44,
                "low": 183.89,
                "close": 185.64,
                "adjusted_close": 185.3,
                "volume": 82488700
            },
            ...
        },
        "news": {
            "AAPL": ["headline 1", "headline 2", ...],
            ...
        },
        "filing_k": {
            "AAPL": "10-K filing text...",
            ...
        },
        "filing_q": {
            "AAPL": "10-Q filing text...",
            ...
        }
    },
    ...
}
```

## CSV Price Data Structure

| Column | Description |
| :--- | :--- |
| `date` | Trading date |
| `symbol` | Ticker symbol |
| `open` | Opening price |
| `high` | Highest price |
| `low` | Lowest price |
| `close` | Closing price |
| `adjusted_close` | Adjusted closing price |
| `volume` | Trading volume |

## Usage

Datasets are auto-downloaded when running FINSABER experiments. See the [FINSABER repo](https://github.com/waylonli/FINSABER) for details.