COPOD: copula-based outlier detection — morie

Per-feature empirical CDFs and their mirrors give tail probabilities; the score is the larger of the summed negative log tail probabilities, with a per-dimension skewness correction. Being rank-based it needs no distance metric and no scaling. Mirrors morie.fn.copod.

Usage

morie_copod(X, skew_correction = TRUE)

Arguments

X: numeric matrix (n x d).
skew_correction: logical; use the skew-selected tail.

Value

list: scores, left_tail, right_tail, skewness, n, d, method.

References

Li, Z., Zhao, Y., Botta, N., Ionescu, C. & Hu, X. (2020). COPOD: copula-based outlier detection. IEEE ICDM, 1118-1123.

Examples

set.seed(1)
X <- matrix(rnorm(90), ncol = 3)
X[1, ] <- 9
which.max(morie_copod(X)$scores)
#> [1] 1