Skip to contents

Calculate features using density-based clustering algorithms

densityBasedFeatures.Rd

The feature_densityBased_all function calculates all of the density-based features in one call. Calculate each feature individually using the feature_densityBased_thetaDiff, feature_densityBased_translationDiff, and feature_densityBased_clusterSize functions. The densityBasedClusters and estimatedRotation functions are exported helpers.

Usage

feature_densityBased_all(
  comparisonData_cellBased,
  eps,
  minPts,
  method = "dbscan",
  id_cols = NULL
)

feature_densityBased_thetaDiff(estimatedThetas, direction, imputeVal = NA)

feature_densityBased_translationDiff(x, y, cluster, direction, imputeVal = NA)

feature_densityBased_clusterSize(cluster, direction, imputeVal = NA)

densityBasedClusters(x, y, eps, minPts, method = "dbscan")

estimatedRotation(x, y, theta)

Arguments

comparisonData_cellBased

tibble such as one returned by the comparison_cellBased() function that contains results from the cell-based comparison procedure

eps

a double representing the neighborhood radius used in the DBSCAN algorithm

minPts

an integer representing the minimum neighborhood size used in the DBSCAN algorithm

method

a character vector that specifies the algorithm used to find clusters. Must be either "dbscan" or "hdbscan" or "optics" - see the dbscan package for more information.

id_cols

column names in the comparisonData tibble that uniquely identify each observation. These are returned along with the computed features

direction

tibble column containing characters "reference_vs_target" or "target_vs_reference" that specify the comparison direction associated with the observation

imputeVal

value to return if the feature calculation results in a non-numeric (i.e., NA, NULL) value

x

tibble column containing estimated horizontal translations

y

tibble column containing estimated vertical translations

theta

tibble column containing estimated rotations

Note

The eps and minPts arguments are used differently depending on the method specified. The eps argument is passed to the eps parameter in the dbscan::dbscan() function if method == "dbscan" and to the eps_cl parameter in the dbscan::extractDBSCAN() function if method == "optics"

See also

dbscan::dbscan(), dbscan::hdbscan(), dbscan::optics() and dbscan::extractDBSCAN()

Examples

data("K013sA1","K013sA2")

compData <- comparison_cellBased(reference = K013sA1,
                                 target = K013sA2,
                                 thetas = c(-3,0,3))
#> Error in mutate(., direction = "reference_vs_target"): could not find function "mutate"

# calculate all density-based features in one call
feature_densityBased_all(compData,eps = 5,minPts = 5)
#> Error in dplyr::group_by(., direction): object 'compData' not found

# verify that we can calculate the same features using the individual
#   feature_densityBased_* functions
compData %>%
  dplyr::group_by(direction) %>%
  dplyr::filter(theta == estimatedRotation(x = x,y = y,theta = theta)) %>%
  dplyr::mutate(clust = densityBasedClusters(x = x,y = y,eps = 5,minPts = 5)) %>%
  dplyr::ungroup() %>%
  dplyr::summarise(thetaDiff = feature_densityBased_thetaDiff(estimatedThetas = theta,direction = direction),
                   translationDiff = feature_densityBased_translationDiff(x = x,y = y,cluster = clust,direction = direction),
                   clusterSize = feature_densityBased_clusterSize(cluster = clust,direction = direction))
#> Error in dplyr::group_by(., direction): object 'compData' not found