ml4t-engineer 0.1.0a3

High-performance quantitative finance feature engineering library

ml4t-engineer

High-performance feature engineering for financial machine learning.

ml4t-engineer provides 107+ technical indicators, triple-barrier labeling, and alternative bar sampling with a Polars-first implementation that's 10-100x faster than pandas alternatives.

Features

• 107+ Technical Indicators: Momentum, trend, volatility, volume, and more
• TA-Lib Validated: 59 indicators validated against TA-Lib at 1e-6 tolerance
• Triple-Barrier Labeling: AFML-compliant labeling with ATR-based barriers
• Alternative Bars: Volume, dollar, tick, and imbalance bars
• Microstructure Metrics: Kyle's Lambda, VPIN, Amihud, Roll spread
• ML-Specific Features: Fractional differencing, entropy, Hurst exponent
• Polars-First: 10-100x faster than pandas, ~0.8x TA-Lib C speed
• Type-Safe: Full type hints with mypy strict mode

Installation

pip install ml4t-engineer

With optional dependencies:

pip install ml4t-engineer[talib]      # TA-Lib backend
pip install ml4t-engineer[numba]      # Numba acceleration
pip install ml4t-engineer[all]        # All optional dependencies

Quick Start

import polars as pl
from ml4t.engineer import compute_features, list_features

# See available features
print(list_features("momentum"))  # RSI, MACD, Stochastic, etc.

# Load OHLCV data
df = pl.read_parquet("ohlcv.parquet")

# Compute features with default parameters
result = compute_features(df, ["rsi", "macd", "atr", "obv"])

# Or with custom parameters
result = compute_features(df, [
    {"name": "rsi", "params": {"period": 20}},
    {"name": "sma", "params": {"period": 50}},
    {"name": "bollinger_bands", "params": {"period": 20, "std_dev": 2.0}},
])

Feature Categories

Category	Count	Examples
Momentum	31	RSI, MACD, Stochastic, CCI, ADX, MFI
Trend	10	SMA, EMA, WMA, DEMA, TEMA, KAMA
Volatility	15	ATR, Bollinger, Yang-Zhang, GARCH
Volume	3	OBV, AD, ADOSC
Statistics	8	Variance, Linear Regression, Correlation
Math	3	MAX, MIN, SUM
Price Transform	5	Typical Price, Weighted Close
Microstructure	12	Kyle Lambda, VPIN, Amihud, Roll
ML	11	Fractional Diff, Entropy, Hurst

Triple-Barrier Labeling

from ml4t.engineer.labeling import triple_barrier_labels, atr_barriers

# Fixed barriers
labels = triple_barrier_labels(
    df,
    upper_barrier=0.02,  # 2% profit target
    lower_barrier=0.01,  # 1% stop loss
    max_holding=20,       # 20 bar horizon
)

# Dynamic ATR-based barriers
labels = atr_barriers(
    df,
    atr_period=14,
    upper_multiplier=2.0,  # 2x ATR profit target
    lower_multiplier=1.0,  # 1x ATR stop loss
    max_holding=20,
)

Alternative Bar Sampling

from ml4t.engineer.bars import volume_bars, dollar_bars, tick_imbalance_bars

# Volume bars (equal volume per bar)
vbars = volume_bars(tick_data, volume_threshold=1000)

# Dollar bars (equal dollar volume per bar)
dbars = dollar_bars(tick_data, dollar_threshold=1_000_000)

# Tick imbalance bars (information-driven)
ibars = tick_imbalance_bars(tick_data, expected_imbalance=100)

Preprocessing

from ml4t.engineer import Preprocessor, StandardScaler, RobustScaler

# Leakage-safe preprocessing
preprocessor = Preprocessor([
    StandardScaler(),
])

# Fit on train only, transform both
X_train_scaled = preprocessor.fit_transform(X_train)
X_test_scaled = preprocessor.transform(X_test)

Configuration via YAML

# features.yaml
features:
  - name: rsi
    params:
      period: 14
  - name: macd
    params:
      fast: 12
      slow: 26
      signal: 9
  - name: bollinger_bands
    params:
      period: 20
      std_dev: 2.0

result = compute_features(df, "features.yaml")

Performance

Benchmark	ml4t-engineer	pandas-ta	Speedup
RSI (1M rows)	12ms	850ms	70x
MACD (1M rows)	18ms	1200ms	67x
Bollinger (1M rows)	15ms	920ms	61x
Triple-barrier (1M rows)	20ms	N/A	-

Benchmarks on M1 MacBook Pro with Polars 0.20+

API Reference

Core Functions

from ml4t.engineer import (
    compute_features,   # Compute features from config
    list_features,      # List available features
    list_categories,    # List feature categories
    describe_feature,   # Get feature metadata
)

Labeling

from ml4t.engineer.labeling import (
    triple_barrier_labels,  # Triple-barrier method
    atr_barriers,           # ATR-based barriers
    meta_labels,            # Meta-labeling
)

Bars

from ml4t.engineer.bars import (
    volume_bars,           # Volume bars
    dollar_bars,           # Dollar bars
    tick_imbalance_bars,   # Tick imbalance bars
    volume_imbalance_bars, # Volume imbalance bars
)

Preprocessing

from ml4t.engineer import (
    Preprocessor,      # Preprocessing pipeline
    StandardScaler,    # Z-score normalization
    RobustScaler,      # Robust scaling (median/IQR)
    MinMaxScaler,      # Min-max scaling
)

Integration with ML4T Libraries

ml4t-engineer is part of the ML4T library ecosystem:

from ml4t.data import DataManager
from ml4t.engineer import compute_features
from ml4t.engineer.labeling import triple_barrier_labels
from ml4t.diagnostic import Evaluator
from ml4t.backtest import Engine

# Complete workflow
data = DataManager().fetch("SPY", "2020-01-01", "2023-12-31")
features = compute_features(data, ["rsi", "macd", "atr"])
labels = triple_barrier_labels(data, 0.02, 0.01, 20)
# ... train model, evaluate, backtest

Development

# Clone repository
git clone https://github.com/applied-ai/ml4t-engineer.git
cd ml4t-engineer

# Install with dev dependencies
uv pip install -e ".[dev]"

# Run tests
uv run pytest tests/ -v

# Type checking
uv run mypy src/

# Linting
uv run ruff check src/

Testing

# Run all tests
uv run pytest tests/

# Run specific test file
uv run pytest tests/test_api.py

# Run with coverage
uv run pytest tests/ --cov=ml4t.engineer

# TA-Lib validation tests (requires TA-Lib)
uv run pytest tests/test_talib_validation.py

References

• López de Prado, M. (2018). Advances in Financial Machine Learning. Wiley.
• López de Prado, M. (2020). Machine Learning for Asset Managers. Cambridge.
• Easley, D., López de Prado, M., & O'Hara, M. (2012). "Flow Toxicity and Liquidity in a High-Frequency World."

License

MIT License - see LICENSE for details.

Contributing

Contributions are welcome! Please read our Contributing Guide for details.