smilegeng/hourly-stock-data-2023

Hourly Stock Prices + Technical Indicators (2023) — TsFile

Converted to Apache TsFile format from the original Hugging Face dataset mdnh/hourly-stock-data-2023. The data introduction below is kept identical to the original dataset card. License: CC-BY-4.0 (same as source).

This dataset contains hourly OHLCV price data and key technical indicators for 8 major U.S. tickers across different sectors. Perfect for time series forecasting, technical analysis, and machine learning projects.

Coverage: January 3, 2023 – December 18, 2023
Symbols: AAPL, MSFT, NVDA, JPM, XOM, SPY, TSLA, AMZN
Records: 11,202
Size: 2.16 MB (original CSV)

---

📊 Columns

Column	Description
timestamp	Date & time in UTC (YYYY-MM-DD HH:MM:SS)
symbol	Stock ticker
open, high, low, close, volume	OHLCV data
sma_10, sma_50	Simple moving averages
ema_20	Exponential moving average
rsi_14	Relative Strength Index
macd, macd_signal, macd_hist	MACD components
volatility_20	Rolling volatility (20-hour window)
target_up_next	Binary target: 1 if next hour close ≥ 0.05% higher

---

⚙️ Technical Details

• Data source: Publicly available financial market data (2023), aggregated and preprocessed to include technical indicators and binary movement labels.
• Interval: 1 hour (aggregated from minute-level data)
• Technical indicators: Calculated using pandas with proper groupby operations per symbol
• Missing values: 16 rows (0.14%) in volatility_20 column - occurs at the start of each symbol's time series where insufficient history exists for 20-hour rolling window
• Timestamps: UTC format, ISO 8601 compliant (YYYY-MM-DD HH:MM:SS)

📈 Data Quality

• ✅ No duplicate records
• ✅ All prices positive and valid
• ✅ All volumes positive
• ✅ Timestamps properly formatted
• ✅ Target variable balanced (41.75% ups, 58.25% downs)

---

🔄 TsFile Conversion

This repository stores the dataset as a single hourly_stock_2023.tsfile (Apache TsFile, the native columnar time-series format of Apache IoTDB).

• No columns were dropped — all 16 source columns are preserved.
• Device dimension (TAG): symbol — each of the 8 tickers is one independent time series.
• Time: the source timestamp (UTC) is stored as the TsFile time column in INT64 milliseconds since the Unix epoch.
• FIELD columns and types:
  • open, high, low, close and the technical indicators sma_10, sma_50, ema_20, rsi_14, macd, macd_signal, macd_hist, volatility_20 → DOUBLE (source float64).
  • volume, target_up_next → INT64 (source int64).
• Missing values: the 16 missing volatility_20 values are kept as null (not imputed, not dropped), matching the original.
• The source has a single CSV (no train/test split), so the output is a single TsFile.

Reading the TsFile

from tsfile import TsFileReader

reader = TsFileReader("hourly_stock_2023.tsfile")
for name, schema in reader.get_all_table_schemas().items():
    print(name, [(c.get_column_name(), c.get_data_type(), c.get_category())
                  for c in schema.get_columns()])

# Query all fields for the table (time column is returned automatically)
cols = [c.get_column_name() for c in schema.get_columns()]
with reader.query_table("hourly_stock_2023", cols, batch_size=65536) as rs:
    while (batch := rs.read_arrow_batch()) is not None:
        ...  # batch is a pyarrow.RecordBatch

---

Attribution

• Original dataset: mdnh/hourly-stock-data-2023
• License: CC-BY-4.0
• This is a format conversion only; the data values are unchanged.