Compute Mutual Information

Computes mutual information between each feature in the features data frame and the target vector. The features are discretized using the "equalfreq" method from infotheo::discretize().

Usage

mutinfo(features, target, return_colnames = FALSE)

Arguments

features

A data frame of features. These will be discretized using the "equalfreq" method (see infotheo::discretize()).

target

A vector (character or factor) representing the variable to compute mutual information with.

return_colnames

Logical; if TRUE, returns the column names from features ordered by their mutual information with target (highest to lowest). If FALSE, returns mutual information values. (default: FALSE)

Value

If return_colnames = FALSE, a named numeric vector of mutual information scores is returned (one for each column in features), sorted in descending order. The names of the vector correspond to the column names of features. If return_colnames = TRUE, only the ordered column names of features are returned.

Examples

data(iris)
# Compute mutual information scores
mutinfo(iris[1:4], iris[[5]])
#>  Petal.Width Petal.Length Sepal.Length  Sepal.Width 
#>    0.8328724    0.8278344    0.4292148    0.2499608 

# Get column names ordered by mutual information with target column (most mutual info first)
mutinfo(iris[1:4], iris[[5]], return_colnames = TRUE)
#> [1] "Petal.Width"  "Petal.Length" "Sepal.Length" "Sepal.Width"