create_CBMS15_attributes: Generate CBMS'15+ Attributes
In unmnn/evoxploit: Augment Evolution Features to a Longitudinal Study Dataset

Description Usage Arguments Value References See Also

Generates CBMS'15+ attributes (evolution features) from a data frame and the output of a clustering with clustering, according to \insertCiteNiemann:CBMS15evoxploit.

create_CBMS15_attributes(
  df,
  label,
  li_clustering,
  train_lgc = rep(TRUE, nrow(df)),
  suffix,
  verbose = FALSE,
  ...
)

`df`	A data frame.
`label`	The target variable, given as either factor or numeric variable.
`li_clustering`	The output of `clustering`.
`suffix`	A string indicating the start of the wave index suffix.
`verbose`	Whether or not to show some diagnostic messages. Defaults to FALSE.

A data frame with the following columns:

dist_s_x_sy: The distance of an instance to itself.
dev_from_pop_a_sx_y: For numeric attributes: The deviation from the population slope for a between x and y. For factors: whether the instance value AND at least 50% of the population have changed.
cluster_idx_sx: The instance's cluster ID.
lof_sx: The instance's Local Outlier Factor \insertCiteBreuning:LOF2000evoxploit.
cluster_rep_sx: Whether the instance is the cluster's representative. The cluster representative is the instance i, where the sum of path distances between i and all cluster peers is minimal.
dist_to_rep_sx: The instance's distance to its cluster representative. NA if noise instance. The distance matrix is inherited from li_clustering.
path_length_to_rep_sx: The path length of direct-density connected instances between an intance and its cluster representative.
dist_to_centroid_sx: The (HEOM) distance between an instance to its centroid.
dist_to_medoid_sx: The (HEOM) distance between an instance to its medoid.
frac_class_c_in_neighborhood_sx: fraction of instances of class c in the instance's ε-neighborhood
silhouette_sx: the instance's silhouette score. (noise instances are treated as one separate cluster.)
real_diff_att_a_sx_y: the real difference of a between y and x for an instance. For numeric attributes only.
abs_diff_att_a_sx_y: the absolute difference of a between y and x for an instance. For numeric attributes only.
rel_diff_att_a_sx_y: the relative difference of a between y and x for an instance. For numeric attributes only.
has_changed_att_a_sx_y: whether the instance has changed in a between x and y. For factors only.
stays_outlier_sx_y: whether the instance was noise in x and remains noise in y.
becomes_outlier_sx_y: whether the instance was not noise in x and becomes noise in y.
was_outlier_sx_y: whether the instance was noise in x and is not noise in y.
never_outlier_sx_y: whether the instance neither was noise in x nor in y.
frac_same_peers_s_x_y: the fraction of cluster peers in both x and y.
same_minPts_nn_s_x_y: the fraction of minPts-nearest neighbors in both x and y.
real_diff_silhouette_s_x_y: Silhouette change.
real_diff_lof_s_x_y: LOF change.
diff_s_x_y: The instance's change w.r.t. the CFS subset of the clustering.