Feature selection with multiple algorithms.
The component includes as many as possible CRAN R packages that do attribute/feature selection. Methods that have minimal parameters are included.
If a package is not installed - it is skipped. If an algorithm fails, or returns an empty set, the output is not written.
Note that due to unsolved installation problems, some of the requirements are set to manual.
Version | 1.0 |
---|---|
Bundle | tools |
Categories | Classification |
Authors | Ville Rantanen (ville.rantanen@helsinki.fi) |
Issue tracker | View/Report issues |
Requires | R ; gsettings-desktop-schemas (DEB) ; propOverlap (R-bioconductor) ; rfUtilities (R-bioconductor) ; RRF (R-bioconductor) ; varSelRF (R-bioconductor) ; AUCRF (R-bioconductor) ; VSURF (R-bioconductor) ; FeaLect (R-bioconductor) ; CORElearn (R-bioconductor) ; e1071 (R-bioconductor) ; Biocomb (R-bioconductor) ; FSelector (R-bioconductor) ; Boruta (R-bioconductor) |
Source files | component.xml AttributeSelection.r |
Usage | Example with default values |
Name | Type | Mandatory | Description |
---|---|---|---|
in | CSV | Mandatory | Input data |
Name | Type | Description |
---|---|---|
out | Array<CSV> | Lists of features with each different method, and Summary with a frequency of the feature. The methods that produce an importance number, include it. |
Name | Type | Default | Description |
---|---|---|---|
classCol | string | (no default) | Column name of the class of each row. Some methods require a binary class problem. If more, those methods are skipped. |
iterations | int | 200 | Number of iterations for methods that require it. |
libraries | string | "AUCRF,Biocomb,Boruta,CORElearn,FeaLect,FSelector,propOverlap,RRF,varSelRF,VSURF" | Comma separated list of libraries to use. |
seed | int | 2015 | Seed for randomization |
skipCol | string | "" | Column names not to use for selection |
Test case | Parameters▼ | IN in |
OUT out |
|||
---|---|---|---|---|---|---|
case1_simple_case | properties | in | out | |||
classCol=Class, |