pymlfinance 0.1.1

Python bindings for the mlfinance AFML toolkit

pymlfinance

Python bindings for mlfinance — a Rust implementation of concepts from Advances in Financial Machine Learning by Marcos López de Prado.

Installation

pip install pymlfinance

With Polars plugin support:

pip install pymlfinance[polars]

Quick Start

import numpy as np
import pymlfinance as ml

# EWMA smoothing
prices = np.array([100.0, 101.5, 99.8, 102.3, 101.0])
smoothed = ml.core.ewma(prices, 3)

# CUSUM event filter
events = ml.data.cusum_filter(prices, 1.5)

# Triple-barrier labeling
from pymlfinance import TripleBarrierConfig
config = TripleBarrierConfig(upper_barrier=0.02, lower_barrier=0.02, max_holding_period=10)

# Fractional differentiation
stationary = ml.sampling.frac_diff_ffd(prices, 0.5, 1e-4)

# HRP portfolio allocation
returns = np.random.randn(100, 5) * 0.02
weights = ml.features.hrp_weights(returns)

# Sharpe ratio
ret = np.random.randn(252) * 0.01
sr = ml.backtesting.sharpe_ratio(ret)

Modules

Module	Description
pymlfinance.core	EWMA, cumsum, returns, matrix utilities
pymlfinance.data	Bar generation (tick, volume, dollar, information-driven), CUSUM filter
pymlfinance.labeling	Triple-barrier, meta-labeling, trend-scanning
pymlfinance.sampling	Sequential bootstrapping, sample weights, fractional differentiation
pymlfinance.features	Structural breaks, entropy, microstructure, HRP, denoising
pymlfinance.modeling	Bet sizing, cross-validation, feature importance
pymlfinance.backtesting	Backtest statistics, overfitting detection, strategy risk

Polars Plugin

import polars as pl
import pymlfinance  # registers .ml namespace

df = pl.DataFrame({"price": [100.0, 101.5, 99.8, 102.3, 101.0]})
df.with_columns(pl.col("price").ml.ewma(span=3).alias("smoothed"))

Documentation

• Python API Documentation
• GitHub Repository

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 — see LICENSE.