Feature Selection in R using glmnet-lasso, xgboost and ranger
This R package wraps glmnet-lasso, xgboost and ranger to perform feature selection. After downloading use ? to read info about each function (i.e. ?feature_selection). More details can be found in the blog-post (http://mlampros.github.io/2016/02/14/feature-selection/). To download the latest version from Github use,
remotes::install_github('mlampros/FeatureSelection')Package Updates:
- Currently there is a new version of glmnet (3.0.0) with new functionality (relax, trace, assess, bigGlm), however it requires an R version of 3.6.0 (see the new vignette for more information).
- In the ranger R package the ranger::importance_pvalues() was added
- Currently, the recommended approach for future selection is SHAP
UPDATE 03-02-2020
Docker images of the FeatureSelection package are available to download from my dockerhub account. The images come with Rstudio and the R-development version (latest) installed. The whole process was tested on Ubuntu 18.04. To pull & run the image do the following,
docker pull mlampros/featureselection:rstudiodev
docker run -d --name rstudio_dev -e USER=rstudio -e PASSWORD=give_here_your_password --rm -p 8787:8787 mlampros/featureselection:rstudiodevThe user can also bind a home directory / folder to the image to use its files by specifying the -v command,
docker run -d --name rstudio_dev -e USER=rstudio -e PASSWORD=give_here_your_password --rm -p 8787:8787 -v /home/YOUR_DIR:/home/rstudio/YOUR_DIR mlampros/featureselection:rstudiodev
In the latter case you might have first give permission privileges for write access to YOUR_DIR directory (not necessarily) using,
The USER defaults to rstudio but you have to give your PASSWORD of preference (see www.rocker-project.org for more information).
Open your web-browser and depending where the docker image was build / run give,
1st. Option on your personal computer,
2nd. Option on a cloud instance,
to access the Rstudio console in order to give your username and password.
