On his Automated Trading blog, Max Dama takes us through the process of implementing a trading model using bagged decision trees in R. Not only does he provide the code and sample data (illustrating how to optimize trading of the financial instrument GLD via backtesting), he also walks us through the code step-by-step in a narrated video. It's a nice illustration of the practical steps of building bespoke models to financial data in R.
The code makes use of the Andy Liaw's randomForest package to fit the bagged decision trees, a process that lends itself nicely to parallel computing. If you have a multiprocessor machine (or a cluster) and REvolution R Enterprise, the randomShrubbery function in ParallelR acts as a drop-in replacement for randomForest and runs in parallel, reducing the computation time for the backtesting.
Max Dama on Automated Trading: Decision Tree Bagging System (R Code)
Google can't find any documentation for randomShubbery - can you help?!
Posted by: Anon | June 03, 2009 at 10:41
randomShrubbery is available with REvolution R Enterprise. It's part of the included ParallelR suite.
Posted by: David Smith | June 03, 2009 at 11:49