In ngloe/olpsR: Algorithms and functions for On-line Portfolio Selection

olpsR - On-line Portfolio Selection with R

The olpsR package provides different On-line Portfolio Selection algorithms and functions to deal with the on-line portfolio selection problem where a portfolio is rebalanced in every period to achieve certain goals, e.g. maximizing terminal wealth. Datasets to test portfolio selection algorithms are also included.

For a background on On-line Portfolio Selection see for example [@LH14; http://arxiv.org/pdf/1212.2129.pdf].

Features

On-Line Portfolio Selection algorithms:

• Buy-and-Hold
• Buy-and-Hold best
• Constant Rebalanced Portfolio
• Best Constant Rebalanced Portfolio
• Universal Portfolio [@Cov91]
• Exponential Gradient [@HSS98]
• Successive Constant Rebalanced Portfolio [@GS00]
• Anticor [@BEG04]
• Passive Aggressive Mean Reversion [@LZH12]
• Confidence Weighted Mean Reversion [@LHZ13]
• Volatility Timing [@KO12]

Functions:

• transform asset prices into returns (price relatives)
• transform returns (price relatives) into asset prices
• calculate portfolio wealth of rebalanced portfolios
• transform a sequence of price relatives into a Kelly market sequence
• projection onto a simplex
• ...

Datasets:

NYSE, DJIA, SP500, TSE, DAX

Installation

To install the olpsR package run:

if (!require("devtools")) install.packages("devtools")
devtools::install_github("ngloe/olpsR")

Getting started

Once installed, the package can be loaded using:

library(olpsR)

To test portfolio selection algorithms some return data is loaded using the NYSE dataset. We select two assets, kinar and iroqu:

library(olpsR)

data(NYSE)
x = cbind(kinar=NYSE$kinar, iroqu=NYSE$iroqu)

Algorithms can be computed by applying alg_ALG on the selected data where ALG is the desired algorithm. For example, to approximate the Universal Portfolio algorithm (UP) type:

UP = alg_UP(x); UP

Accessing UP then returns a short summary of the algorithm's output. To access the calculated portfolio wealth or the portfolio weights you can type:

UP$Wealth
UP$Weights

The achieved portfolio wealth (performance) can be plotted by:

plot(UP)

To easily compare different algorithms pass them to the plot function:

BH_Market = alg_BH( x, weights=c(0.5, 0.5) )
BH_best   = alg_BHbest(x)
plot(BH_Market, BH_best, UP)

Getting help

For more details and an overview of the implemented algorithms and functions please refer to the package help by typing:

help(package="olpsR")

References

ngloe/olpsR documentation built on May 23, 2019, 4:42 p.m.