PSGExpress (version 3.1.3)

PSGExpress-package: PSGExpress

Description

PSGExpress

Details

Portfolio Safeguard (PSG) is a decision-support tool for solving a wide range of optimization, statistics, and risk management problems. PSG includes four classes of optimization solvers intended for optimizing linear, nonlinear (may be non-smooth), and mixed-integer problems. Special attention is paid to problems involving uncertainties in performance functions (such as Variance, VaR, CVaR, drawdown, bPOE).

Sets of nonlinear functions are identified in various engineering areas and these functions are precoded. Functions are independent objects with defined data structure, list of variables, and output. Codes for solving optimization problems are very simple, just several lines. On the other hand, this approach allows for building specialized very fast algorithms for various classes of functions and applications.

Although PSG is a general-purpose decision support tool, the focus applications areas are risk management, financial engineering, statistics, logistics, and medical applications. Financial applications are especially well covered, such as portfolio optimization, asset allocation, selection of insurance contracts, hedging with derivative contracts, bond matching, and structuring of Collateralized Debt Obligations (CDO). Case studies are posted at these links:

<http://aorda.com/html/PSG_CS_HTML/index.html?case_studies_full_list.htm>,

<https://www.ise.ufl.edu/uryasev/research/>.

See Zabarankin and Uryasev (2014) <doi:10.1007/978-1-4614-8471-4>.

Examples

# NOT RUN {
#Problem of CVaR minimization with constraint on the mean profit:
#Find x = (x1,x2,x3,x4) minimizing
#risk(x) = CVaR(0.95,x)
#subject to
#Average Gain(x)>4.5
#x1+x2+x3+x4 = 1
#x1>=0, x2>=0, x3>=0, x4>=0

matrix_scenarios <- matrix(c(1,4,8,3, 7,5,4,6, 2,8,1,0,0,3,4,9),nrow=4, byrow=TRUE)
colnames(matrix_scenarios) <- colnames(matrix_scenarios,do.NULL = FALSE, prefix = "x")
scenario_benchmark <- c(0.2, 0.11, 0.6, 0.1)
matrix_scenarios <- cbind(matrix_scenarios,scenario_benchmark)
matrix_budget <- matrix(c(1, 1, 1, 1),nrow=1)
colnames(matrix_budget) <- colnames(matrix_budget,do.NULL = FALSE, prefix = "x")
point_lowerbounds <- c(0, 0, 0, 0)
names(point_lowerbounds) <- rownames(point_lowerbounds,do.NULL = FALSE, prefix = "x")

# Case 1. PSG data is saved in list
problem_list <- list()

#Problem Statement
problem_list$problem_statement <- sprintf(
 "minimize
 cvar_risk(0.95,matrix_scenarios)
Constraint: >= 4.5
 avg_g(matrix_scenarios)
Constraint: == 1
 linear(matrix_budget)
Box: >= point_lowerbounds")

# PSG Matrix:
problem_list$matrix_scenarios <- matrix_scenarios

# PSG Matrix:
problem_list$matrix_budget <- matrix_budget

# PSG Point:
problem_list$point_lowerbounds <- point_lowerbounds

# Solve optimization problem
output.list.1 <- rpsg_solver(problem_list)

print(output.list.1)

# Case 2. PSG data is saved in Global Environment
problem_list <- list()

#Problem Statement
problem_list$problem_statement <- sprintf(
 "minimize
 cvar_risk(0.95,matrix_scenarios)
Constraint: >= 4.5
 avg_g(matrix_scenarios)
Constraint: == 1
 linear(matrix_budget)
Box: >= point_lowerbounds")

output.list.2 <- rpsg_solver(problem_list)

print(output.list.2)
# }