Crate rust_portfolio_opt

Expand description

The crate mirrors PyPortfolioOpt’s module structure:

expected_returns — historical mean / EMA / CAPM return estimators.
risk_models — sample / EWMA / semi-covariance + Ledoit-Wolf and oracle-approximating shrinkage; cov ↔ corr helpers.
efficient_frontier — mean-variance optimisation: minimum variance, tangency (max Sharpe), efficient risk / return, with weight bounds.
black_litterman — equilibrium prior + view-blended posterior.
hrp — hierarchical risk parity (correlation distance, single linkage, recursive bisection).
cla — Markowitz’s Critical Line Algorithm.
discrete_allocation — convert continuous weights to integer share counts under a budget.

All matrix / vector inputs and outputs use nalgebra’s DMatrix / DVector so they slot into broader nalgebra-based pipelines.

Each estimator and optimiser also has a _labeled companion that accepts ticker names alongside the prices and returns BTreeMap<String, f64> (ordered by ticker), mirroring how PyPortfolioOpt accepts a pandas.DataFrame and returns an OrderedDict. See LabeledVector / LabeledMatrix for the intermediate types.

Every estimator and optimiser keeps the same default annualisation convention as PyPortfolioOpt: 252 trading days unless explicitly overridden by the caller.

Modules§

black_litterman: Black-Litterman: blend an equilibrium prior with investor views.
cla: Critical Line Algorithm (CLA) for exact mean-variance efficient frontier tracing.
discrete_allocation: Discrete allocation: convert continuous portfolio weights into integer share counts given a total budget.
efficient_frontier: Mean-variance optimisation à la pypfopt.efficient_frontier.
expected_returns: Expected-return estimators.
hrp: Hierarchical Risk Parity (López de Prado, 2016).
risk_models: Covariance / risk-matrix estimators.

Structs§

LabeledMatrix: A square N×N matrix paired with N ticker labels (used for both row and column indexing — the same labelling pandas applies to a covariance/correlation DataFrame).
LabeledVector: A length-N vector of floats paired with N ticker labels. Mirrors a pandas Series indexed by ticker name. Returned by every estimator that has a _labeled companion (e.g. crate::expected_returns::mean_historical_return_labeled) so callers can keep ticker order alongside the numerical values.

Enums§

PortfolioError: Crate-wide error type. Most fallible operations return PortfolioError rather than panicking so callers can decide how to recover from misconfigured inputs (mismatched shapes, infeasible optimisation, etc.).

Constants§

TRADING_DAYS_PER_YEAR: Default annualisation factor — 252 trading days per year, matching PyPortfolioOpt’s defaults.

Functions§

clean_weights: Round entries of weights whose absolute value is below cutoff to zero, renormalise the remainder so they sum to the original total, and optionally round to rounding decimal places. Mirrors PyPortfolioOpt’s base_optimizer.BaseOptimizer.clean_weights.
column_means: Column-wise mean of an T x N matrix → length-N vector.
log_returns_from_prices: Log returns: ln(p_t / p_{t-1}).
returns_from_prices: Compute simple period-over-period returns from a T x N price matrix.
sample_covariance: Sample covariance with Bessel’s correction (divisor T-1).
symmetrise: Symmetrise a matrix in-place ((A + A^T) / 2). Useful after numerical operations that produce minute asymmetry.
to_weight_map: Convert a (values, tickers) pair into a ticker-keyed BTreeMap, matching PyPortfolioOpt’s habit of returning weights as an OrderedDict[str, float]. Errors if the lengths disagree.

Type Aliases§

Result: Convenience alias used by every module in the crate.