Skip to contents

Find and prune highly correlated variables

Source: R/fs_correlation.R
fs_correlation.Rd

Finds pairs of variables whose absolute correlation exceeds threshold and, by default, drops variables until no such pair remains, keeping one representative of each correlated group.

Usage

fs_correlation(
  data,
  threshold,
  method = "pearson",
  prune = TRUE,
  na.rm = FALSE,
  sample_frac = 1,
  output_format = "matrix",
  diag_value = 0,
  seed = NULL,
  verbose = FALSE,
  parallel = FALSE,
  n_cores = 2L
)

Arguments

data

A data frame or matrix with at least 2 columns. Column names must be unique, non-empty, and not NA; duplicated names are an error because a correlation matrix with duplicated names is ambiguous. The allowed column types depend on method:

  • "pearson", "spearman", "kendall": all numeric.

  • "polychoric": all ordered factors.

  • "pointbiserial": numeric, or any type with exactly 2 distinct non-missing values. A numeric column with more than 2 distinct values counts as continuous, and a column with exactly 2 counts as dichotomous. At least one of each is needed, or every correlation is NA (with a warning).

threshold

Absolute-correlation cutoff between 0 and 1. Pairs with an absolute correlation strictly greater than threshold are flagged. Required; there is no default.

method

One of "pearson" (default), "spearman", "kendall", "polychoric" (needs the suggested polycor package), or "pointbiserial". The point-biserial correlation is the Pearson correlation between a continuous variable and a 0/1 indicator of a dichotomous one, where 1 marks the second sorted value (for a factor, the second level). Only continuous-dichotomous pairs have one, so under this method every continuous-continuous and dichotomous-dichotomous cell stays NA and those pairs are never flagged.

prune

If TRUE (default), selected is the reduced set left after pruning (see "How pruning works"). If FALSE, nothing is dropped and selected is every variable in at least one flagged pair: the redundant set, not the set to keep.

na.rm

How to handle missing values. If TRUE, they are removed pairwise for "pearson", "spearman", and "kendall"; within each pair for "pointbiserial"; and casewise (rows complete across all columns) for "polychoric". If FALSE (default), a missing value makes the affected correlations NA, except under "polychoric", which stops with an error instead of silently dropping rows.

sample_frac

Fraction of rows, in (0, 1], to sample without replacement before computing correlations. The sample size is rounded up and is never below one row. Default 1 (no sampling).

output_format

Shape of details$corr_matrix: "matrix" (default) or "data.frame" (long form).

diag_value

Single number, or NA, written on the diagonal of details$corr_matrix. The diagonal never takes part in flagging, scoring, or pruning. Default 0.

seed

Optional integer seed for the row sampling. It is applied for this call only and the previous RNG state is restored afterward. It has no effect when sample_frac = 1, and parallel point-biserial computation uses no random numbers. Default NULL (featR does not touch the RNG).

verbose

If TRUE, report row sampling, the correlation method, and any variables dropped by pruning. Default FALSE.

parallel

If TRUE, compute point-biserial correlations in parallel with the suggested foreach and doParallel packages. Every other method computes the matrix in one call and ignores this. Default FALSE.

n_cores

Number of workers when parallel = TRUE and method = "pointbiserial", a whole number of at least 1, capped at the detected core count. Default 2.

Value

An fs_result object with:

selected

With prune = TRUE, every variable not in details$dropped; no two of them have a computable absolute correlation above threshold. With prune = FALSE, both members of every flagged pair.

scores

Named numeric vector giving each variable's largest absolute correlation with any other variable. Larger means more redundant. NA when no correlation with that variable could be computed.

method

paste0("correlation_", method), for example "correlation_pearson".

task

NA; the filter uses no outcome.

model

NULL; no model is fitted.

details

A list with:

  • corr_matrix: the correlation matrix with diag_value on the diagonal, or, when output_format = "data.frame", the same values in long form with columns Var1, Var2, Correlation.

  • pairs: a data frame of the flagged pairs (Var1, Var2, Correlation), strongest absolute correlation first, whatever output_format is.

  • dropped: variables removed by pruning, in the order they were dropped; empty when prune = FALSE.

  • redundant: every variable in at least one flagged pair, which is selected as it would be under prune = FALSE.

  • n_features: the number of variables considered.

call

The matched call.

Details

This is an unsupervised redundancy filter. It looks only at how the variables relate to one another, never at an outcome, so when two variables are nearly interchangeable it cannot keep the one that predicts better. It keeps whichever is less correlated with the rest of the data. Use fs_boruta() when the choice within a correlated group should depend on importance for a target.

How pruning works

Pruning is greedy. While any pair of retained variables still has an absolute correlation above threshold, the strongest such pair is taken, and the member with the larger mean absolute correlation to the other retained variables is dropped (on an exact tie, the later column). Afterward no two retained variables have a computable correlation above threshold. The surviving set is not guaranteed to be the smallest one with that property, and it can depend on the order in which pairs are resolved.

This borrows the tie-break of caret::findCorrelation() but not its scan order: caret visits columns in order of mean correlation rather than pairs in order of strength, so the two can drop different variables. Variables that were never flagged are always kept.

Correlations that cannot be computed

A correlation that cannot be computed is NA. Such pairs are never flagged and never pruned, because an unknown correlation is no evidence of redundancy, and a warning reports how many there are. When no pair exceeds threshold, a message (not a warning) says so, whatever verbose is set to.

See also

The Filter methods guide, which runs this function on a worked example.

Examples

d <- data.frame(
  a = c(1, 2, 3, 4, 5, 6),
  b = c(2, 4, 6, 8, 10, 12),
  c = c(1.5, 0.9, 2.1, 0.4, 1.1, 0.8)
)
res <- fs_correlation(d, threshold = 0.9)
res$selected
#> [1] "a" "c"
res$details$pairs
#>   Var1 Var2 Correlation
#> 1    a    b           1

# Both members of every flagged pair, with nothing dropped
fs_correlation(d, threshold = 0.9, prune = FALSE)$selected
#> [1] "a" "b"