pymlfinance¶

Python bindings for the mlfinance AFML toolkit — a high-performance Rust implementation of the methods from Advances in Financial Machine Learning by Marcos López de Prado.

Overview¶

pymlfinance provides ~140 functions and 20+ classes across 7 modules, covering:

Core — Statistical primitives: EWMA, returns, correlation/covariance matrices, eigendecomposition
Data — 10 bar aggregation methods (tick, volume, dollar, imbalance, runs) plus sampling and ETF trick
Labeling — Triple-barrier method, meta-labeling, volatility estimators, trend scanning
Sampling — Fractional differentiation, sequential bootstrap, concurrency, sample weights
Features — Structural breaks (SADF/GSADF), entropy, microstructure, RMT denoising, portfolio allocation (HRP/CLA/IVP), clustering (ONC), codependence measures
Modeling — Purged K-fold CV, feature importance (MDI/MDA/SFI), hyperparameter search, scoring
Backtesting — Performance statistics (Sharpe, PSR, DSR), overfitting detection (PBO/CSCV), strategy risk, bet sizing, synthetic data generation

All computation happens in Rust with zero-copy NumPy array interchange via PyO3. A native Polars expression plugin is also included for use in Polars pipelines (see Polars Integration).

Quick Example¶

import numpy as np
import pymlfinance as ml

# Generate log returns
prices = np.array([100.0, 101.5, 99.8, 102.3, 101.0, 103.5])
returns = ml.core.log_returns(prices)

# Compute daily volatility with EWMA
vol = ml.core.ewma_std(returns, span=3)

# Fractional differentiation to preserve memory while achieving stationarity
min_d = ml.sampling.find_min_d(prices, p_value_threshold=0.05, max_d=1.0)
stationary = ml.sampling.frac_diff_ffd(prices, d=min_d, threshold=1e-4)

# Sharpe ratio
sr = ml.backtesting.sharpe_ratio(returns, periods_per_year=252)

Installation¶

# From source (requires Rust toolchain)
cd crates/pymlfinance
pip install maturin
maturin develop --release

Disclaimer¶

This software is provided for educational and research purposes only. It does not constitute financial advice. Use of this code for trading or investment decisions is entirely at your own risk. The authors accept no liability for any financial losses incurred.

License¶

BUSL-1.1