Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/mlr-org/mlr3filters
Filter-based feature selection for mlr3
https://github.com/mlr-org/mlr3filters
feature-selection filter filters mlr mlr3 r r-package variable-importance
Last synced: 1 day ago
JSON representation
Filter-based feature selection for mlr3
- Host: GitHub
- URL: https://github.com/mlr-org/mlr3filters
- Owner: mlr-org
- License: lgpl-3.0
- Created: 2018-10-17T12:48:36.000Z (about 6 years ago)
- Default Branch: main
- Last Pushed: 2024-09-24T10:28:31.000Z (about 1 month ago)
- Last Synced: 2024-10-14T04:48:03.360Z (25 days ago)
- Topics: feature-selection, filter, filters, mlr, mlr3, r, r-package, variable-importance
- Language: R
- Homepage: https://mlr3filters.mlr-org.com
- Size: 10.1 MB
- Stars: 20
- Watchers: 11
- Forks: 8
- Open Issues: 11
-
Metadata Files:
- Readme: README.Rmd
- Changelog: NEWS.md
- License: LICENSE
Awesome Lists containing this project
- jimsghstars - mlr-org/mlr3filters - Filter-based feature selection for mlr3 (R)
README
---
output: github_document
---
# mlr3filters
Package website: [release](https://mlr3filters.mlr-org.com/) | [dev](https://mlr3filters.mlr-org.com/dev/)
{mlr3filters} adds feature selection filters to [mlr3](https://mlr3.mlr-org.com).
The implemented filters can be used stand-alone, or as part of a machine learning pipeline in combination with
[mlr3pipelines](https://mlr3pipelines.mlr-org.com) and the [filter operator](https://mlr3pipelines.mlr-org.com/reference/mlr_pipeops_filter.html).
Wrapper methods for feature selection are implemented in [mlr3fselect](https://mlr3fselect.mlr-org.com).
Learners which support the extraction feature importance scores can be combined with a filter from this package for embedded feature selection.
[](https://github.com/mlr-org/mlr3filters/actions/workflows/r-cmd-check.yml)
[](https://cran.r-project.org/package=mlr3filters)
[](https://stackoverflow.com/questions/tagged/mlr3)
[](https://lmmisld-lmu-stats-slds.srv.mwn.de/mlr_invite/)
## Installation
CRAN version
```{r eval = FALSE}
install.packages("mlr3filters")
```
Development version
```{r, eval = FALSE}
remotes::install_github("mlr-org/mlr3filters")
```
## Filters
### Filter Example
```{r}
set.seed(1)
library("mlr3")
library("mlr3filters")
task = tsk("sonar")
filter = flt("auc")
head(as.data.table(filter$calculate(task)))
```
### Implemented Filters
```{r echo = FALSE, message=FALSE}
library("mlr3misc")
library("mlr3filters")
library("data.table")
link_cran = function(pkg) {
mlr3misc::map(pkg, function(.x) {
mlr3misc::map_chr(.x, function(.y) {
if (unlist(.y) %in% getOption("defaultPackages")) {
.y
} else {
sprintf("[%1$s](https://cran.r-project.org/package=%1$s)", .y)
}
})
})
}
tab = as.data.table(mlr_filters)[, !c("params", "task_properties")]
tab[, task_types := sapply(task_types, function(x) if (is_scalar_na(x)) "Universal" else paste(capitalize(x), collapse = " & "))]
tab[, feature_types := sapply(feature_types, function(x) paste(capitalize(x), collapse = ", "))]
tab[, packages := sapply(packages, function(x) paste(link_cran(x), collapse = ", "))]
# manually change the task type for specific filters
learner_based = c("performance", "permutation", "importance", "selected_features")
tab[key %in% learner_based, task_types := "Universal"]
tab[key %in% learner_based, packages := ""]
setnames(tab,
old = c("key", "task_types", "feature_types", "packages"),
new = c("Name", "Task Types", "Feature Types", "Package")
)
knitr::kable(tab, format = "markdown")
```
### Variable Importance Filters
The following learners allow the extraction of variable importance and therefore are supported by `FilterImportance`:
```{r echo=FALSE, warning=FALSE}
library("mlr3learners")
tab = as.data.table(mlr_learners)
tab[sapply(properties, is.element, el = "importance"), key]
```
If your learner is not listed here but capable of extracting variable importance from the fitted model, the reason is most likely that it is not yet integrated in the package [mlr3learners](https://github.com/mlr-org/mlr3learners) or the [extra learner extension](https://github.com/mlr-org/mlr3extralearners).
Please open an issue so we can add your package.
Some learners need to have their variable importance measure "activated" during learner creation.
For example, to use the "impurity" measure of Random Forest via the {ranger} package:
```{r}
task = tsk("iris")
lrn = lrn("classif.ranger", seed = 42)
lrn$param_set$values = list(importance = "impurity")
filter = flt("importance", learner = lrn)
filter$calculate(task)
head(as.data.table(filter), 3)
```
### Performance Filter
`FilterPerformance` is a univariate filter method which calls `resample()` with every predictor variable in the dataset and ranks the final outcome using the supplied measure.
Any learner can be passed to this filter with `classif.rpart` being the default.
Of course, also regression learners can be passed if the task is of type "regr".
### Filter-based Feature Selection
In many cases filtering is only one step in the modeling pipeline.
To select features based on filter values, one can use [`PipeOpFilter`](https://mlr3pipelines.mlr-org.com/reference/mlr_pipeops_filter.html) from [mlr3pipelines](https://github.com/mlr-org/mlr3pipelines).
```{r, results='hide'}
library(mlr3pipelines)
task = tsk("spam")
# the `filter.frac` should be tuned
graph = po("filter", filter = flt("auc"), filter.frac = 0.5) %>>%
po("learner", lrn("classif.rpart"))
learner = as_learner(graph)
rr = resample(task, learner, rsmp("holdout"))
```