PortOpt {strand}R Documentation

Portfolio optimization class

Description

The PortOpt object is used to set up and solve a portfolio optimization problem.

Details

A PortOpt object is configured in the same way as a Simulation object, by supplying configuration in a yaml file or list to the object constructor. Methods are available for adding constraints and retrieving information about the optimization setup and results. See the package vignette for information on configuration file setup.

Methods

Public methods

Method new()

Create a new PortOpt object.

Usage
PortOpt$new(config, input_data)
Arguments
config

An object of class list or character. If the value passed is a character vector, it should be of length 1 and specify the path to a yaml configuration file that contains the object's configuration info. If the value passed is of class list(), the list should contain the object's configuration info in list form (e.g, the return value of calling yaml.load_file on the configuration file).

input_data

A data.frame that contains all necessary input for the optimization.

If the top-level configuration item price_var is not set, prices will be expected in the ref_price column of input_data.

Returns

A new PortOpt object.

Examples
library(dplyr)
data(sample_secref)
data(sample_inputs)
data(sample_pricing)

# Construct optimization input for one day from sample data. The columns
# of the input data must match the input configuration.
optim_input <-
  inner_join(sample_inputs, sample_pricing,
             by = c("id", "date")) %>%
  left_join(sample_secref, by = "id") %>%
  filter(date %in% as.Date("2020-06-01")) %>%
  mutate(ref_price = price_unadj,
                shares_strategy_1 = 0)

opt <-
  PortOpt$new(config = example_strategy_config(),
              input_data = optim_input)

# The problem is not solved until the \code{solve} method is called
# explicitly.
opt$solve()

Method setVerbose()

Set the verbose flag to control the amount of informational output.

Usage
PortOpt$setVerbose(verbose)
Arguments
verbose

Logical flag indicating whether to be verbose or not.

Returns

No return value, called for side effects.

Method addConstraints()

Add optimization constraints.

Usage
PortOpt$addConstraints(constraint_matrix, dir, rhs, name)
Arguments
constraint_matrix

Matrix with one row per constraint and (S+1) \times N columns, where S is number of strategies and N is the number of stocks.

The variables in the optimization are

x_{1,1}, x_{2,1}, \ldots, x_{N,1},

x_{1,2}, x_{2,2}, \ldots, x_{N,2},

\vdots

x_{1,S}, x_{2,S}, \ldots, x_{N,S},

y_1, \ldots, y_N

The first N \times S variables are the individual strategy trades. Variable x_{i,s} represents the signed trade for stock i in strategy s. The following N auxillary variables y_1, \ldots, y_N represent the absolute value of the net trade in each stock. So for a stock i, we have:

y_i = \sum_s |x_{i,s}|

dir

Vector of class character of length nrow(constraint_matrix) that specifies the direction of the constraints. All elements must be one of ">=", "==", or "<=".

rhs

Vector of class numeric of length nrow(constraint_matrix) that specifies the bounds of the constraints.

name

Character vector of length 1 that specifies a name for the set of constraints that are being created.

Returns

No return value, called for side effects.

Method getConstraintMatrix()

Constraint matrix access.

Usage
PortOpt$getConstraintMatrix()
Returns

The optimization's constraint matrix.

Method getConstraintMeta()

Provide high-level constraint information.

Usage
PortOpt$getConstraintMeta()
Returns

A data frame that contains constraint metadata, such as current constraint value and whether a constraint is currently within bounds, for all single-row constraints. Explicitly exclude net trade constraints and constraints that involve net trade variables.

Method solve()

Solve the optimization. After running solve(), results can be retrieved using getResultData().

Usage
PortOpt$solve()
Returns

No return value, called for side effects.

Method getResultData()

Get optimization result.

Usage
PortOpt$getResultData()
Returns

A data frame that contains the number of shares and the net market value of the trades at the strategy and joint (net) level for each stock in the optimization's input.

Method getLoosenedConstraints()

Provide information about any constraints that were loosened in order to solve the optimization.

Usage
PortOpt$getLoosenedConstraints()
Returns

Object of class list where keys are the names of the loosened constraints and values are how much they were loosened toward current values. Values are expressed as (current constraint value - loosened constraint value) / (current constraint value - violated constraint value). A value of 0 means a constraint was loosened 100% and is not binding.

Method getMaxPosition()

Provide information about the maximum position size allowed for long and short positions.

Usage
PortOpt$getMaxPosition()
Returns

An object of class data.frame that contains the limits on size for long and short positions for each strategy and security. The columns in the data frame are:

id

Security identifier.

strategy

Strategy name.

max_pos_lmv

Maximum net market value for a long position.

max_pos_smv

Maximum net market value for a short position.

Method getMaxOrder()

Provide information about the maximum order size allowed for each security and strategy.

Usage
PortOpt$getMaxOrder()
Returns

An object of class data.frame that contains the limit on order size for each strategy and security. The columns in the data frame are:

id

Security identifier.

strategy

Strategy name.

max_order_gmv

Maximum gross market value allowed for an order.

Method summaryDf()

Provide aggregate level optimization information if the problem has been solved.

Usage
PortOpt$summaryDf()
Returns

A data frame with one row per strategy, including the joint (net) level, and columns for starting and ending market values and factor expoure values.

Method print()

Print summary information.

Usage
PortOpt$print()
Returns

No return value, called for side effects.

Method clone()

The objects of this class are cloneable with this method.

Usage
PortOpt$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Examples


## ------------------------------------------------
## Method `PortOpt$new`
## ------------------------------------------------

library(dplyr)
data(sample_secref)
data(sample_inputs)
data(sample_pricing)

# Construct optimization input for one day from sample data. The columns
# of the input data must match the input configuration.
optim_input <-
  inner_join(sample_inputs, sample_pricing,
             by = c("id", "date")) %>%
  left_join(sample_secref, by = "id") %>%
  filter(date %in% as.Date("2020-06-01")) %>%
  mutate(ref_price = price_unadj,
                shares_strategy_1 = 0)

opt <-
  PortOpt$new(config = example_strategy_config(),
              input_data = optim_input)

# The problem is not solved until the \code{solve} method is called
# explicitly.
opt$solve()
[Package strand version 0.2.0 Index]