Has anyone that uses R for submissions outlined or written about their setup? I’m an R specialist and prefer working in R rather than Python. Also, if there are any other R nerds out there, 
2 Likes
I fall into the category of R nerds. I use mostly R/Rstudio + Rnumerai library. My approach is a set of “croned” R scripts that in a few steps generate predictions: these basic steps: get data, make some extra features & auxiliary targets (with neutralization), estimate a set of models (with main and auxiliary targets), combine models with different optimization metrics in mind (like Sharpe, Sortino, etc.) and submit predictions.
awesome! what does “croned” mean in this case?
edit : “scheduled under cron” would make sense now that i think of it
task that is running the script on schedule:
in my win10 case it’s some *.bat file that contain only a few strings
@echo on
"c:\Program Files\R\R-3.6.1\bin\R.exe" CMD BATCH d:\2020\numerai2020\R\run.start.A.ssh.v04.r
You could also look into the compute service provided by numerai. Using a python package it sets up an AWS service that runs a docker image (costs can be as low as 1 USD per month depending on the set-up and runtime - mine is 2 USD). In the image you can use R code. I have used this set-up for a year now. It works very well. For submitting you can use the Rnumerai package. If you need more info then I come back in the next couple of days and edit this post with more details.
1 Like
thanks! I should be able to figure it out, though feel free if you want to!