Tutorial 01: Real Data and DeTime Feature Factory

Tutorial 01 builds the reusable data and feature layer for the series.

Live data path

from quant_trading.data import fetch_yahoo_ohlcv

ohlcv = fetch_yahoo_ohlcv("SPY", start="2016-01-01")

Offline real-data smoke path

from quant_trading.data import load_sample_goog_ohlcv

ohlcv = load_sample_goog_ohlcv(trim_start="2010-01-01")

The bundled fallback is real historical GOOG OHLCV data from the uploaded Learn Algorithmic Trading archive. It is included so the notebooks and smoke tests run without network access. It is not a substitute for a licensed research data vendor.

Feature factory

from quant_trading.decomposition_features import walkforward_price_volume_features

prices = ohlcv[["Close"]].rename(columns={"Close": "SPY"})
volumes = ohlcv[["Volume"]].rename(columns={"Volume": "SPY"})

features = walkforward_price_volume_features(
    prices,
    volumes,
    method="STL",
    period="auto",
    period_candidates=(63, 126, 252),
    train_window=504,
    step=5,
    z_window=63,
)

Returned feature names include:

Feature	Meaning
trend, trend_slope, trend_strength	price direction and persistence
cycle, cycle_slope, cycle_z, cycle_amplitude	oscillatory timing and amplitude
residual_z, residual_abs_z, residual_vol	structural deviation and stress
volume_trend_slope, volume_residual_z, volume_participation	participation and abnormal activity
component_stability, reconstruction_error	reliability screens

Executable notebook: examples/notebooks/quant_trading/01_market_data_and_decomposition_feature_factory.ipynb.