README

Overview

PortfolioAnalytics is a comprehensive R package designed to provide numerical methods and visualization tools for portfolio optimization and analysis. It is built to handle complex portfolio optimization problems with a wide range of constraints and objectives, supporting both traditional and modern risk measures.

Main Use Cases

• Constrained Optimization: Define and solve optimization problems with box, group, turnover, diversification, leverage, and factor exposure constraints.
• Custom Objectives: Optimize for various targets including mean return, variance, expected shortfall (CVaR), risk budgets, and coherent second moments (CSM).
• Backtesting: Evaluate portfolio performance over time with rebalancing and walk-forward analysis.
• Visualization: Analyze efficient frontiers, risk-reward trade-offs, and weight distributions through extensive charting capabilities.

Advanced Features

• CVXR Integration: Seamlessly use the CVXR solver infrastructure for convex optimization problems (LP, QP, SOCP, SDP, MIP).
• Robust Covariance Estimators: Compute outliers-robust portfolios using MM-estimators, Rocke's estimator, Minimum Covariance Determinant (MCD), and more.
• Multi-layer Optimization: Define hierarchical portfolio structures for sub-portfolio optimization.
• Regime Switching: Optimize portfolios across different market regimes.
• Custom Moments and Objectives: Flexibility to provide user-defined moment functions and objective functions.

Supported Optimization Engines

PortfolioAnalytics supports a variety of global and local optimization engines:

• CVXR: A modern convex optimization interface supporting various solvers like ECOS, OSQP, and SCS.
• Random Portfolios: Fast, heuristic optimization using sample, simplex, or grid-based portfolio generation.
• DEoptim: Differential Evolution optimization for non-convex and non-smooth problems.
• ROI (R Optimization Infrastructure): A unified interface to multiple solvers including GLPK, Symphony, and Quadprog.
• GenSA: Generalized Simulated Annealing for global optimization.
• pso: Particle Swarm Optimization.
• mco: Multi-objective optimization using genetic algorithms (e.g., NSGA-II).
• osqp: Operator Splitting Solver for Quadratic Programming.
• Rglpk: Interface to the GNU Linear Programming Kit.

Documentation and Examples

Comprehensive documentation is available in the vignettes directory:

• portfolio_vignette: Introductory guide to the package.
• cvxrPortfolioAnalytics: Deep dive into using CVXR for convex optimization.
• robustCovMatForPA: Detailed explanation of robust covariance matrix estimators.
• ROI_vignette: Guide to using the ROI solver infrastructure.
• risk_budget_optimization: Focus on risk budgeting techniques.
• custom_moments_objectives: Instructions for extending the package with custom logic.

The package also includes extensive examples and demo scripts:

• demo/: Over 30 scripts covering everything from basic setup to replicating research papers (e.g., demo_JPM2024MinDownsideRisk.R).
• inst/examples/: Additional example scripts for specific use cases.

Release Summary

[2.1.x] - Current

• Enhanced Visualization: Improved graphical displays for multiple efficient frontiers and robust covariance estimator settings.
• Solver Updates: Continued refinements to CVXR and ROI integrations.
• JPM 2024 Paper Replication: Added scripts to replicate "Minimum Downside Risk Portfolios" (October 2024).
• Risk Measure Update: Replaced EQS (Expected Quadratic Shortfall) with CSM (Coherent Second Moment).
• Test Suite Expansion: Added comprehensive test suite with the aid of Gemini and Claude.

[2.0.x] - 2024

• CVXR Integration: Major feature allowing access to eleven solver packages via CVXR.
• MCSM Portfolios: Capability for Minimum Coherent Second Moment (MCSM) portfolios using SOCP.
• Robust Portfolio Optimization: Integrated several robust covariance matrix estimators (MM, Rocke, Mcd, TSGS).
• New Utility Functions: Added extract_risk, chart.EfficientFrontierCompare, opt.outputMvo, and more.

[1.x] - 2013-2023

• Rebalancing Support: Enhanced optimize.portfolio.rebalancing for backtesting and walk-forward analysis.
• Expanded Statistics: Added extractStats and refined S3 methods for optimization outputs.
• Group Constraints: Robust support for hierarchical group-based constraints.

[0.x] - 2009-2012

• Package Foundation: Initial development (originally named optimizer).
• Core Solvers: Introduced DEoptim and Random Portfolios support.
• Initial ROI Support: Early integration with the R Optimization Infrastructure.

Thanks

PortfolioAnalytics has benefited immensely from the support of the open-source community and academic funding:

• Google Summer of Code (GSoC): We are grateful for the project funding provided by Google over several years, including 2013, 2014, and 2022.
• Student Contributors:
  • Xinran Zhao (GSoC 2022) - CVXR integration and CSM risk measures.
  • Yifu Kang (GSoC 2022) - Robust covariance matrix estimators.
  • Ross Bennett (GSoC 2013, 2014) - Core framework development, rebalancing, and documentation.
• Mentors: Doug Martin, Steve Murray, Brian Peterson, Peter Carl, and Guy Yollin.
• Community: Thank you to all contributors who have submitted bug reports, feature requests, and code patches. Special thanks to the many users who have shared their expertise to make this package a industry standard for portfolio optimization in R.

For bug reporting and contributions, please visit our GitHub Issues page.