In jeanmarcgp/xtsanalytics: A library of functions to do advanced analytics of time series data (xts matrices).

The xtsanalytics package provides a set of function to analyze and develop investment strategies, including:

• General purpose functions for data analysis and manipulation of xts objects
• Several types of xts plotting and visualization functions
• Features and indicator generating functions, suitable for build machine learning models
• Functions to explore the autocorrelation of returns, including of higher moments
• A random forest walk-forward optimization system to make better asset returns predictions

The package also leverages and extends several finance related packages. These packages should be installed since xtsanalytics requires them.

statfolio data structure and Walk-Forward Optimization

xtsanalytics makes extensive use of the xts data structure. In addition, a new data structure is created to easily build statistical equity curves of a portfolio using Walk-Forward optimization (WFO). This is called the statfolio data structure.
There may be different classes of statfolio depending on which function was used to generate it. For example, function optimize_statfolio() creates a statfolio of class "optimize.portfolio.rebalancing".

WFO optimizes an equity curve at a series of specific points in time using a rolling window. For example, a typical use would be to optimize monthly, at the end of each month. In doing so, the data window used is N days up to and including the WFO date, resulting in an equity curve up to the current WFO date. The following month, the same is repeated to generate the section of equity curve from the previous WFO date up to and including the current WFO date. For example, for a 10 year equity curve, this would result in 10 years * 12 months = 120 WFO optimizations if done monthly. Each of these blocks of time (between two consecutive WFO dates) can be used to create a subfolio.

A subfolio consists of multiple optimizations (or models) performed between two consecutive WFO dates. For instance, a portfolio optimizer (see optimize_statfolio) may converge only to an approximate set of portfolio weights. In such a case, it may be prudent to run several optimizations with different seeds in order to get a set of weights. The average of these weights could then be used in a real portfolio. In addition, the distribution of these weights can be analyzed to see how much variance exists between these optimization results, and reach conclusions regarding the importance of rebalancing to accurate weights. This set of multiple weights is the subfolio.

The statfolio is defined to be the set of all subfolios resulting from a WFO run with more than a single optimization performed as each WFO dates.

Using statfolios to generate equity curve statistics

A statfolio can be used to generate equity curve statistics. For example, let's assume that wfo_statfolio() is run on a set of assets, with argument N = 5. Setting N = 5 means that the optimization is repeated 5 times at each WFO dates. Let's further assume monthly optization.

If the WFO optimization is done over 12 months, there will be 5^12 = 244 million possible equity curve paths, since for every month, there are 5 possible paths to choose! Function stitch_statfolio() can be used to generate a large number of such randomly picked equity curve paths to see the statistical effects on the equity curve resulting from the optimizer approximations.

Core functions available in xtsanalytics

• xtsplot() is an easy to use plotting function. It allows multiple equity curves with a nice default set of colors, normalization of curves at the beginning, log scales, shaded regions and several other useful visualization features.
• make_features() is used to generate a set of features from a matrix of asset prices. This is useful for a machine learning model for instance. Each feature is an xts matrix, and these are returned in a list.
• xtsoverlay() is used to align time series to a common point in time (t0). This is used to help visualize (using xtsplot) how curves can line up when time-shifted. Returns a zoo matrix instead of xts since the date index is lost, as the index is now relative.
• wfo_grow_forest() is used to perform walk-forward optimization on a single set of parameter using the random forest machine learning algorithm. It is single-threaded.
• wfo_grow_jungle() calls wfo_grow_forest() multiple times, each with a different set of parameters. It is set up to use multiple threads using package doParallel. Since doParallel does not allow to print to the console, a live report is provided in a log file to monitor progress.

Analysis and general purpose functions

• cumprod_na() is used to build an equity curve from returns
• interpolate_returns() can be used to interpolate an equity curve when merging with another one with finer time granularity. This is useful, for example, when plotting a monthly curve overlaid with a daily curve.
• starttime() is used to start a timer before doing a CPU-intensive task.
• elapsedtime() is used to report the elapsed time since starttime() was last invoked.

Vignette Info

Note the various macros within the vignette setion of the metadata block above. These are required in order to instruct R how to build the vignette. Note that you should change the title field and the \VignetteIndexEntry to match the title of your vignette.

Styles

The html_vignette template includes a basic CSS theme. To override this theme you can specify your own CSS in the document metadata as follows:

output: 
  rmarkdown::html_vignette:
    css: mystyles.css

Figures

The figure sizes have been customised so that you can easily put two images side-by-side.

#plot(1:10)
#plot(10:1)

You can enable figure captions by fig_caption: yes in YAML:

output:
  rmarkdown::html_vignette:
    fig_caption: yes

Then you can use the chunk option fig.cap = "Your figure caption." in knitr.

More Examples

You can write math expressions, e.g. $Y = X\beta + \epsilon$, footnotes^[A footnote here.], and tables, e.g. using knitr::kable().

#knitr::kable(head(mtcars, 10))

Also a quote using >:

"He who gives up [code] safety for [code] speed deserves neither." (via)

jeanmarcgp/xtsanalytics documentation built on May 19, 2019, 12:38 a.m.