Score how much each point looks like an outlier

Returns the local outlier factor of every point in the array, one score per row.

Usage

ga_outlier_scores(geometry, k_neighbours)

Arguments

geometry: a GeoArrow point array
k_neighbours: how many neighbours define a neighbourhood

Value

a double array of scores, the same length as geometry

Details

A score near 1 means a point sits at the same density as its neighbours. Scores meaningfully above 1 mean it is in a sparser neighbourhood than they are, which is what marks an outlier. There is no universal cutoff, so compare scores within a dataset rather than against a fixed threshold.

Use k_neighbours to set how many neighbours define the local neighbourhood. With small values ga_outlier_scores() reacts to fine structure, and with large values it smooths that structure away. It scores over the whole array, not within each row, so k_neighbours is a single value. It returns null for a row that is not a single point, or is null.

References

OutlierDetection

Examples

g <- ga_point(
  c(0, 0.1, 0.2, 0.3, 9),
  c(0, 0.1, 0.2, 0.3, 9)
)

round(as.vector(ga_outlier_scores(g, k_neighbours = 2)), 2)
#> [1]    1    1    1    1 3785