The R packages oneClass implements the one-class classifiers one-class SVM, biased SVM, and Maxent, as custom functions for the train function of the package caret. Thus, the extensive infrastructure of caret can be used for training and analyzing one-class classification models. The infrastructure is further extended by one-class classification specific tools which may help to understand and thus improve one-class classifier outcomes in the absence of a representative and complete test data. The package is developed for one-class land cover classification with remote sensing data. Therefore, a part of classical data frames it also contains methods to predict raster data.
The most important functionalities are described in the introductory vignette.
Note that the package is still in developement and needs further testing. Also some documentation might still be incomplete.
If you encounter a bug, unclear/lacking documentation or any other problem with the package please use the issue tracker or contact me.
The package is available on GitHub. You can install it from within R with the package devtools and the following command:
install_github('benmack/oneClass')
You can develop and use the oneClass source code in a Docker container build from the benmack/r-oneclass-deps image. Prerequisites are that you have Docker installed and local copy of the oneClass source code on your machine.
Run the following command in a terminal:
docker run -i -p 8787:8787 -e PASSWORD=<password of your choice> -v <path containing the local copy of the oneClass repository>:/home/rstudio/oneClass benmack/r-oneclass-deps:3.5.3
And go to http://localhost:8787/ in your browser where you can log in with the user rstudio and the password of your choice to connect to RStudio. In RStudio you should find the oneClass folder under the files in which you can start the oneClass.Rproj.
Note that the Dockerimage is optimized and for not as slim as it could be and that maxent.jar is not yet included in the image (see https://hub.docker.com/r/benmack/r-oneclass-deps).
The following package looks like a very interesting future base image on top of which oneClass might be installed without much of a setup cost because it contains most if not all dependencies, including maxent:
docker run -rm -p 8888:8888 -e JUPYTER_ENABLE_LAB=yes -v ${PWD}:/home/jovyan/work scioquiver/notebooks:cgspatial-notebook