| Title: | Feature Importance for Partitional Clustering |
| Version: | 0.1.5 |
| Description: | Implements a novel approach for measuring feature importance in k-means clustering. Importance of a feature is measured by the misclassification rate relative to the baseline cluster assignment due to a random permutation of feature values. An explanation of permutation feature importance in general can be found here: https://christophm.github.io/interpretable-ml-book/feature-importance.html. |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| Suggests: | flexclust, clustMixType, knitr, rmarkdown, testthat, attempt, ClustImpute, covr |
| Imports: | ggplot2 |
| RoxygenNote: | 7.1.1 |
| Depends: | data.table |
| NeedsCompilation: | no |
| Packaged: | 2021-10-20 16:04:14 UTC; opfaf |
| Author: | Oliver Pfaffel [aut, cre] |
| Maintainer: | Oliver Pfaffel <opfaffel@gmail.com> |
| Repository: | CRAN |
| Date/Publication: | 2021-10-20 16:20:02 UTC |
Feature importance for k-means clustering
Description
This function loops through PermMisClassRate for each variable of the data.
The mean misclassification rate over all iterations is interpreted as variable importance.
Usage
FeatureImpCluster(
clusterObj,
data,
basePred = NULL,
predFUN = NULL,
sub = 1,
biter = 10
)
Arguments
clusterObj |
a "typical" cluster object. The only requirement is that there must be a prediction function which maps the data to an integer |
data |
data.table with the same features as the data set used for clustering (or the simply the same data) |
basePred |
should be equal to results of predFUN(clusterObj,newdata=data); this option saves time when data is a very large data set |
predFUN |
predFUN(clusterObj,newdata=data) should provide the cluster assignment as a numeric vector; typically this is a wrapper around a build-in prediction function |
sub |
integer between 0 and 1(=default), indicates that only a subset of the data should be used if <1 |
biter |
the permutation is iterated biter(=5, default) times |
Value
A list of
- misClassRate
A matrix of the permutation misclassification rate for each variable and each iteration
- featureImp
For each row of complete_data, the associated cluster
Examples
set.seed(123)
dat <- create_random_data(n=1e3)$data # random data
library(flexclust)
res <- kcca(dat,k=4)
f <- FeatureImpCluster(res,dat)
plot(f)
Permutation misclassification rate for single variable
Description
Answers the following question: Using the current partion as a baseline, what is the misclassification rate if a given feature is permuted?
Usage
PermMisClassRate(
clusterObj,
data,
varName,
basePred = NULL,
predFUN = NULL,
sub = 1,
biter = 5,
seed = 123
)
Arguments
clusterObj |
a "typical" cluster object. The only requirement is that there must be a prediction function which maps the data to an integer |
data |
data.table with the same features as the data set used for clustering (or the simply the same data) |
varName |
character; variable name |
basePred |
should be equal to results of predFUN(clusterObj,newdata=data); this option saves time when data is a very large data set |
predFUN |
predFUN(clusterObj,newdata=data) should provide the cluster assignment as a numeric vector; typically this is a wrapper around a build-in prediction function |
sub |
integer between 0 and 1(=default), indicates that only a subset of the data should be used if <1 |
biter |
the permutation is iterated biter(=5, default) times |
seed |
value for random seed |
Value
vector of length biter with the misclassification rate
Examples
set.seed(123)
dat <- create_random_data(n=1e3)$data # random data
library(flexclust)
res <- kcca(dat,k=4)
PermMisClassRate(res,dat,varName="x")
Create random data set with 4 clusters
Description
Create random data set with 4 clusters in a 2 dimensional subspace of a nr_other_vars+2 dimensional space
Usage
create_random_data(n = 10000, nr_other_vars = 4)
Arguments
n |
number of points |
nr_other_vars |
number of other variables / "noise" dimensions |
Value
list containing the random data.table and a vector with the true underlying cluster assignments
Examples
create_random_data(n=1e3)
Feature importance box plot
Description
Feature importance box plot
Usage
## S3 method for class 'featImpCluster'
plot(x, dat = NULL, color = "none", showPoints = FALSE, ...)
Arguments
x |
an object returned from FeatureImpCluster |
dat |
same data as used for the computation of the feature importance (only relevant for colored plots) |
color |
If set to "type", the plot will show different variable types with a different color. |
showPoints |
Show points (default is False) |
... |
arguments to be passed to base plot method |
Value
Returns a ggplot2 object