quant-pml 0.1.0

No project description provided

Quantitative Portfolio Machine Learning Library

Viacheslav Buchkov
ETH Zürich

Keywords: quant; quantitative finance; portfolio management; machine learning; reinforcement learning.

Install

pip install quant-pml

How To Run A Backtest

from quant_pml.config.trading_config import TradingConfig
from quant_pml.config.russell_3000_experiment_config import Russell3000ExperimentConfig
from quant_pml.hedge.market_futures_hedge import MarketFuturesHedge
from quant_pml.runner import build_backtest
from quant_pml.strategies.factors.momentum import Momentum

# Specify the model hyperparameters
REBAL_FREQ = "D" # Rebalance frequency (Month End)
QUANTILE = 0.1 # Sorted portfolio quantile
MODE = "long_short" # Trading mode

# Specify the trading configurations
trading_config = TradingConfig(
    broker_fee=0, # Broker commission in decimals
    bid_ask_spread=0, # Bid-ask spread in decimals
    total_exposure=1, # Budget constraint
    max_exposure=None, # Maximum weight constraint
    min_exposure=None, # Minimum weight constraint
    trading_lag_days=1, # Trading Lag
)

# Create a strategy with logic for weights selection
strategy = Momentum(
    mode=MODE,
    quantile=QUANTILE,
)

# Initialize the hedger (can be set to `None` if no hedging is required)
hedger = MarketFuturesHedge(market_name="spx")

# Build the backtest pipeline
preprocessor, runner = build_backtest(
    experiment_config=Russell3000ExperimentConfig(),
    trading_config=trading_config,
    rebal_freq=REBAL_FREQ,
)

# Run the backtest
res = runner(
    feature_processor=preprocessor,
    strategy=strategy,
    hedger=hedger,
)

How To Create A New Strategy

import pandas as pd
from quant_pml.strategies.base_strategy import BaseStrategy
from quant_pml.strategies.optimization_data import TrainingData, PredictionData


# Create a blueprint for your strategy
class NewStrategy(BaseStrategy):
    def __init__(self) -> None:
        super().__init__()
        
        # Store the data across time, if needed

    def _fit(self, training_data: TrainingData) -> None:
        # Specify here your fitting logic
        pass

    def _get_weights(self, prediction_data: PredictionData, weights_: pd.DataFrame) -> pd.DataFrame:
        # Specify here your prediction logic
        
        # Update the weights DataFrame by modifying the respective columns of stock_ids
        return weights_

How To Create A New Covariance Estimator

from __future__ import annotations

from typing import TYPE_CHECKING

if TYPE_CHECKING:
    import pandas as pd

from quant_pml.strategies.optimization_data import PredictionData, TrainingData
from quant_pml.cov_estimators.base_cov_estimator import BaseCovEstimator


class NewCovEstimator(BaseCovEstimator):
    def __init__(self) -> None:
        super().__init__()

        self._fitted_cov = None

    def _fit(self, training_data: TrainingData) -> None:
        # Specify here your fitting logic (e.g., store a historical covmat)
        ...

    def _predict(self, prediction_data: PredictionData) -> pd.DataFrame:
        # Specify here your prediction logic (e.g., get a stored estimate)
        ...

How To Create A New Data-Driven Covariance Estimator

from __future__ import annotations

from typing import TYPE_CHECKING

if TYPE_CHECKING:
    import pandas as pd

from quant_pml.cov_estimators.rl.base_rl_estimator import BaseRLCovEstimator


class NewRLCovEstimator(BaseRLCovEstimator):
    def __init__(self, shrinkage_type: str, window_size: int | None = None) -> None:
        super().__init__(shrinkage_type=shrinkage_type, window_size=window_size)

    def _fit_shrinkage(
        self, features: pd.DataFrame, shrinkage_target: pd.Series
    ) -> None:
        # Specify here your model (might be classical ML / RL / etc.)
        self.model = ...
        self.model.fit(X=features, y=shrinkage_target)

    def _predict_shrinkage(self, features: pd.DataFrame) -> float:
        pred = self.model.predict(features).item()

        return pred