How skfolio/skfolio works
skfolio is a scikit-learn-compatible Python library for portfolio optimization, factor modeling, and risk management. Users supply asset-return or panel data, optionally transform or pre-select it, estimate priors and risk inputs, then fit optimizers and predict portfolios; model-selection tools validate and compare those workflows. Portfolio objects support downstream performance measurement and attribution. The README also documents distributions, uncertainty sets, synthetic stress scenarios, and factor-model capabilities, represented as subsystem nodes where source details were not sampled. Relationships among unsampled subsystems are intentionally omitted; documented capabilities are identified from the README.
Main entry points: optimization/_base.py Optimizers · _portfolio.py Portfolio results · _asset_panel/_base.py Asset panel · _characteristics_factor_model.py Factor model
Components 22
Data and features
- Asset panel
src/skfolio/containers/_asset_panel/_base.py - Return transforms
src/skfolio/preprocessing - Asset selection
src/skfolio/pre_selection - Descriptors
src/skfolio/descriptor/_base.py
Risk and priors
- Factor exposures
src/skfolio/factor_exposure - Factor model
src/skfolio/prior/_characteristics_factor_model.py - Prior estimators
src/skfolio/prior/_empirical.py - Moment estimators
src/skfolio/moments - Distributions
src/skfolio/distribution - Uncertainty sets
src/skfolio/uncertainty_set
Portfolio allocation
- Optimizers
src/skfolio/optimization/_base.py - Convex methods
src/skfolio/optimization/convex/_base.py - Hierarchical methods
src/skfolio/optimization/cluster/hierarchical - Portfolio results
src/skfolio/portfolio/_portfolio.py
Validation and analysis
- Model selection
src/skfolio/model_selection/_validation.py - Online search
src/skfolio/model_selection/_online/_search.py - Risk measures
src/skfolio/measures/_measures.py - Factor attribution
src/skfolio/attribution/_realized.py - Forecast evaluation
src/skfolio/alpha/_evaluation.py - Portfolio analysis
src/skfolio/population/_population.py
Other components
- Analyst
- Asset data
Connections 25
- Analyst → Asset data: provides returns
- Asset data → Asset panel: organizes data
- Asset panel → Return transforms: transforms data
- Asset panel → Asset selection: selects assets
- Asset panel → Descriptors: computes features
- Descriptors → Factor exposures: informs exposures
- Factor exposures → Factor model: supplies exposures
src/skfolio/prior/_characteristics_factor_model.py - Factor model → Prior estimators: uses factor prior
src/skfolio/prior/_characteristics_factor_model.py - Prior estimators → Moment estimators: estimates moments
- Prior estimators → Distributions: builds distributions
- Uncertainty sets → Optimizers: constrains uncertainty
- Prior estimators → Optimizers: provides prior
src/skfolio/optimization/_base.py - Optimizers → Convex methods: specializes
src/skfolio/optimization/convex/_base.py - Optimizers → Hierarchical methods: specializes
- Optimizers → Portfolio results: predicts portfolios
src/skfolio/optimization/_base.py - Analyst → Model selection: validates models
- Model selection → Optimizers: fits estimators
src/skfolio/model_selection/_validation.py - Model selection → Portfolio results: collects results
src/skfolio/model_selection/_validation.py - Online search → Model selection: uses validation
src/skfolio/model_selection/_online/_search.py - Portfolio results → Risk measures: computes measures
src/skfolio/portfolio/_portfolio.py - Portfolio results → Factor attribution: supports attribution
src/skfolio/portfolio/_portfolio.py - Factor model → Factor attribution: enables factor analysis
- Factor model → Forecast evaluation: evaluates forecasts
- Portfolio results → Portfolio analysis: aggregates portfolios
src/skfolio/population/_population.py - Risk measures → Portfolio analysis: scores portfolios