importance_dtw_cpp: (C++) Contribution of Individual Variables to the...

In distantia: Advanced Toolset for Efficient Time Series Dissimilarity Analysis

(C++) Contribution of Individual Variables to the Dissimilarity Between Two Time Series (Robust Version)

Description

Computes the contribution of individual variables to the similarity/dissimilarity between two irregular multivariate time series. In opposition to the legacy version, importance computation is performed taking the least-cost path of the whole sequence as reference. This operation makes the importance scores of individual variables fully comparable. This function generates a data frame with the following columns:

• variable: name of the individual variable for which the importance is being computed, from the column names of the arguments x and y.

• psi: global dissimilarity score psi of the two time series.

• psi_only_with: dissimilarity between x and y computed from the given variable alone.

• psi_without: dissimilarity between x and y computed from all other variables.

• psi_difference: difference between psi_only_with and psi_without.

• importance: contribution of the variable to the similarity/dissimilarity between x and y, computed as (psi_difference * 100) / psi_all. Positive scores represent contribution to dissimilarity, while negative scores represent contribution to similarity.

Usage

importance_dtw_cpp(
  x,
  y,
  distance = "euclidean",
  diagonal = TRUE,
  weighted = TRUE,
  ignore_blocks = FALSE,
  bandwidth = 1
)

Arguments

x	(required, numeric matrix) multivariate time series.
y	(required, numeric matrix) multivariate time series with the same number of columns as 'x'.
distance	(optional, character string) distance name from the "names" column of the dataset distances (see distances$name). Default: "euclidean".
diagonal	(optional, logical). If TRUE, diagonals are included in the computation of the cost matrix. Default: TRUE.
weighted	(optional, logical). Only relevant when diagonal is TRUE. When TRUE, diagonal cost is weighted by y factor of 1.414214 (square root of 2). Default: TRUE.
ignore_blocks	(optional, logical). If TRUE, blocks of consecutive path coordinates are trimmed to avoid inflating the psi distance. Default: FALSE.
bandwidth	(required, numeric) Size of the Sakoe-Chiba band at both sides of the diagonal used to constrain the least cost path. Expressed as a fraction of the number of matrix rows and columns. Unrestricted by default. Default: 1

Value

data frame

See Also

Other Rcpp_importance: importance_dtw_legacy_cpp(), importance_ls_cpp()

Examples

#simulate two regular time series
x <- zoo_simulate(
  seed = 1,
  rows = 100
  )

y <- zoo_simulate(
  seed = 2,
  rows = 150
  )

#different number of rows
#this is not a requirement though!
nrow(x) == nrow(y)

#compute importance
df <- importance_dtw_cpp(
  x = x,
  y = y,
  distance = "euclidean"
)

df

distantia documentation built on April 4, 2025, 5:42 a.m.