Description Optimization Charts and Graphs Package Dependencies Further Work Acknowledgements Author(s) References See Also
PortfolioAnalytics is an R package to provide numerical solutions for portfolio problems with complex constraints and objective sets. The goal of the package is to aid practicioners and researchers in solving portfolio optimization problems with complex constraints and objectives that mirror real-world applications.
One of the goals of the packages is to provide a common interface to specify constraints and objectives that can be solved by any supported solver (i.e. optimization method). Currently supported optimization methods include
random portfolios
differential evolution
particle swarm optimization
generalized simulated annealing
linear and quadratic programming routines
The solver can be specified with the optimize_method
argument in optimize.portfolio
and optimize.portfolio.rebalancing
. The optimize_method
argument must be one of "random", "DEoptim", "pso", "GenSA", "ROI", "quadprog", "glpk", or "symphony".
Additional information on random portfolios is provided below. The differential evolution algorithm is implemented via the DEoptim package, the particle swarm optimization algorithm via the pso package, the generalized simulated annealing via the GenSA package, and linear and quadratic programming are implemented via the ROI package which acts as an interface to the Rglpk, Rsymphony, and quadprog packages.
A key strength of PortfolioAnalytics is the generalization of constraints and objectives that can be solved.
If optimize_method="ROI"
is specified, a default solver will be selected based on the optimization problem. The glpk
solver is the default solver for LP and MILP optimization problems. The quadprog
solver is the default solver for QP optimization problems. For example, optimize_method = "quadprog"
can be specified and the optimization problem will be solved via ROI using the quadprog plugin package.
The extension to ROI solves a limited type of convex optimization problems:
Maxmimize portfolio return subject leverage, box, group, position limit, target mean return, and/or factor exposure constraints on weights.
Minimize portfolio variance subject to leverage, box, group, turnover, and/or factor exposure constraints (otherwise known as global minimum variance portfolio).
Minimize portfolio variance subject to leverage, box, group, and/or factor exposure constraints and a desired portfolio return.
Maximize quadratic utility subject to leverage, box, group, target mean return, turnover, and/or factor exposure constraints and risk aversion parameter.
(The risk aversion parameter is passed into optimize.portfolio
as an added argument to the portfolio
object).
Maximize portfolio mean return per unit standard deviation (i.e. the Sharpe Ratio) can be done by specifying maxSR=TRUE
in optimize.portfolio
.
If both mean and StdDev are specified as objective names, the default action is to maximize quadratic utility, therefore maxSR=TRUE
must be specified to maximize Sharpe Ratio.
Minimize portfolio ES/ETL/CVaR optimization subject to leverage, box, group, position limit, target mean return, and/or factor exposure constraints and target portfolio return.
Maximize portfolio mean return per unit ES/ETL/CVaR (i.e. the STARR Ratio) can be done by specifying maxSTARR=TRUE
in optimize.portfolio
.
If both mean and ES/ETL/CVaR are specified as objective names, the default action is to maximize mean return per unit ES/ETL/CVaR.
These problems also support a weight_concentration objective where concentration of weights as measured by HHI is added as a penalty term to the quadratic objective.
Because these convex optimization problem are standardized, there is no need for a penalty term. The multiplier
argument in add.objective
passed into the complete constraint object are ingnored by the ROI solver.
Many real-world portfolio optimization problems are global optimization problems, and therefore are not suitable for linear or quadratic programming routines. PortfolioAnalytics provides a random portfolio optimization method and also utilizes the R packages DEoptim, pso, and GenSA for solving non-convex global optimization problems.
PortfolioAnalytics supports three methods of generating random portfolios.
The sample method to generate random portfolios is based on an idea by Pat Burns. This is the most flexible method, but also the slowest, and can generate portfolios to satisfy leverage, box, group, and position limit constraints.
The simplex method to generate random portfolios is based on a paper by W. T. Shaw. The simplex method is useful to generate random portfolios with the full investment constraint (where the sum of the weights is equal to 1) and min box constraints. Values for min_sum and max_sum of the leverage constraint will be ignored, the sum of weights will equal 1. All other constraints such as the box constraint max, group and position limit constraints will be handled by elimination. If the constraints are very restrictive, this may result in very few feasible portfolios remaining. Another key point to note is that the solution may not be along the vertexes depending on the objective. For example, a risk budget objective will likely place the portfolio somewhere on the interior.
The grid method to generate random portfolios is based on the gridSearch
function in package NMOF. The grid search method only satisfies the min and max box constraints. The min_sum and max_sum leverage constraint will likely be violated and the weights in the random portfolios should be normalized. Normalization may cause the box constraints to be violated and will be penalized in constrained_objective
.
PortfolioAnalytics leverages the PerformanceAnalytics package for many common objective functions. The objective types in PortfolioAnalytics are designed to be used with PerformanceAnalytics functions, but any user supplied valid R function can be used as an objective.
This summary attempts to provide an overview of how to construct a portfolio object with constraints and objectives, run the optimization, and chart the results.
The portfolio object is initialized with the portfolio.spec
function. The main argument to portfolio.spec
is assets
. The assets
argument can be a scalar value for the number of assets, a character vector of fund names, or a named vector of initial weights.
Adding constraints to the portfolio object is done with add.constraint
. The add.constraint
function is the main interface for adding and/or updating constraints to the portfolio object. This function allows the user to specify the portfolio to add the constraints to, the type of constraints, arguments for the constraint, and whether or not to enable the constraint. If updating an existing constraint, the indexnum
argument can be specified.
Objectives can be added to the portfolio object with add.objective
. The add.objective
function is the main function for adding and/or updating objectives to the portfolio object. This function allows the user to specify the portfolio to add the objectives to, the type, name of the objective function, arguments to the objective function, and whether or not to enable the objective. If updating an existing objective, the indexnum
argument can be specified.
With the constraints and objectives specified in the portfolio object, the portfolio object can be passed to optimize.portfolio
or optimize.portfolio.rebalancing
to run the optimization. Arguments to optimize.portfolio
include asset returns, the portfolio obect specifying constraints and objectives, optimization method, and other parameters specific to the solver. optimize.portfolio.rebalancing
adds support for backtesting portfolio optimization through time with rebalancing or rolling periods.
Intuition into the optimization can be aided through visualization. The goal of creating the charts is to provide visualization tools for optimal portfolios regardless of the chosen optimization method.
chart.Weights
plots the weights of the optimal portfolio. chart.RiskReward
plots the optimal portfolio in risk-reward space. The random portfolios, DEoptim, and pso solvers will return trace portfolio information at each iteration when optimize.portfolio
is run with trace=TRUE
. If this is the case, chart.RiskReward
will plot these portfolios so that the feasible space can be easily visualized. Although the GenSA and ROI solvers do not return trace portfolio information, random portfolios can be be generated with the argument rp=TRUE
in chart.RiskReward
. A plot
function is provided that will plot the weights and risk-reward scatter chart. The component risk contribution can be charted for portfolio optimization problems with risk budget objectives with chart.RiskBudget
. Neighbor portfolios can be plotted in chart.RiskBudget
, chart.Weights
, and chart.RiskReward
.
Efficient frontiers can be extracted from optimize.portfolio
objects or created from a portfolio
object. The efficient frontier can be charted in risk-reward space with chart.EfficientFrontier
. The weights along the efficient frontier can be charted with chart.Weights.EF
.
Multiple objects created via optimize.portfolio
can be combined with combine.optimizations
for visual comparison. The weights of the optimal portfolios can be plotted with chart.Weights
. The optimal portfolios can be compared in risk-reward space with chart.RiskReward
. The portfolio component risk contributions of the multiple optimal portfolios can be plotted with chart.RiskBudget
.
Several of the functions in the PortfolioAnalytics package require time series data of returns and the xts
package is used for working with time series data.
The PerformanceAnalytics package is used for many common objective functions. The objective types in PortfolioAnalytics are designed to be used with PerformanceAnalytics functions such as StdDev
, VaR
, and ES
.
The foreach and iterators packages are used extensively throughout the package to support parallel programming. The primary functions where foreach
loops are used is optimize.portfolio
, optimize.portfolio.rebalancing
, and create.EfficientFrontier
.
In addition to a random portfolios optimzation method, PortfolioAnalytics supports backend solvers by leveraging the following packages: DEoptim, pso, GenSA, ROI and associated ROI plugin packages.
Continued work to improved charts and graphs.
Continued work to improve features to combine and compare multiple optimal portfolio objects.
Support for more solvers.
Comments, suggestions, and/or code patches are welcome.
TODO
Kris Boudt
Peter Carl
Brian G. Peterson
Maintainer: Brian G. Peterson brian@braverock.com
Shaw, William Thornton, Portfolio Optimization for VAR, CVaR, Omega and Utility with General Return Distributions: A Monte Carlo Approach for Long-Only and Bounded Short Portfolios with Optional Robustness and a Simplified Approach to Covariance Matching (June 1, 2011). Available at SSRN: http://ssrn.com/abstract=1856476 or http://dx.doi.org/10.2139/ssrn.1856476
Scherer, B. and Martin, D. Modern Portfolio Optimization. Springer. 2005.
CRAN task view on Empirical Finance
http://cran.r-project.org/src/contrib/Views/Econometrics.html
CRAN task view on Optimization
http://cran.r-project.org/web/views/Optimization.html
Large-scale portfolio optimization with DEoptim
http://cran.r-project.org/web/packages/DEoptim/vignettes/DEoptimPortfolioOptimization.pdf
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.