aggregate_imp: Aggregate importances

In kerntools: Kernel Functions and Tools for Machine Learning Applications

Aggregate importances

Description

'aggregate_imp()' sums the importances present in a matrix or data.frame according to some user-specified grouping criterion.

Usage

aggregate_imp(X, lev = NULL, samples = "rows")

Arguments

X	Matrix or data.frame containing the importances (in rows or in columns).
lev	(optional) The grouping elements. 'lev' should be as long as the dimension (cols or rows) that one wants to aggregate. If this parameter is absent, the colnames (if samples="rows") or rownames will be used to that effect. In that case, it is expected that the col/rownames follow this pattern: "V_Y", and the variables with the same "V" will be summed. (Check the colnames of a typical output of 'dummy_data()' for more info).
samples	Samples are in rows or in columns? (Defaults: "rows").

Value

X, a matrix or data.frame containing the aggregated importances.

Examples

importances <- matrix(rnorm(90),nrow=3,ncol=30)
rownames(importances) <- c("sample1","sample2","sample3")
colnames(importances) <- paste0("Feat",
rep(1:5,times=2*(1:5)), "_", unlist(lapply(2*(1:5),function(x)LETTERS[1:x])))

## The grouping criterion is:
groups <- paste0("Feat",1:5)
aggregate_imp(X=importances,samples="rows",lev=groups)
## We can also use the colnames:
colnames(importances)
aggregate_imp(X=importances,samples="rows")

kerntools documentation built on April 3, 2025, 7:52 p.m.