pam_cent function

Centroid for partition around medoids

Extract the medoid time series based on a distance measure.

pam_cent(
  series,
  distance,
  ids = seq_along(series),
  distmat = NULL,
  ...,
  error.check = TRUE
)

Arguments

• series: The time series in one of the formats accepted by tslist().
• distance: A character indicating which distance to use. Only needed if distmat is NULL. The distance must be registered in proxy::pr_DB().
• ids: Integer vector indicating which of the series should be considered.
• distmat: Optionally, a pre-computed cross-distance matrix of all series.
• ...: Any extra parameters for the distance function that may be used.
• error.check: Logical indicating whether the function should try to detect inconsistencies and give more informative errors messages. Also used internally to avoid repeating checks.

Returns

The medoid time series.

Details

The medoid's index is determined by taking the $arg min$ of the distmat's row-sums (considering only the rows in ids). The distance matrix is calculated if needed.

Examples

# Computes the distance matrix for all series
pam_cent(CharTraj, "dtw_basic", ids = 6L:10L, window.size = 15L) # series_id = 7L

# Computes the distance matrix for the chosen subset only
pam_cent(CharTraj[6L:10L], "dtw_basic", window.size = 15L) # series_id = 2L