filter_candidate_motifs: Filter Candidate Motifs

In funMoDisco: Motif Discovery in Functional Data

Filter Candidate Motifs

Description

Filters the candidate motifs based on a specified threshold for the average silhouette index and a threshold for the size of the curves in the motif. This function is useful for refining the set of candidate motifs by removing those that do not meet the defined criteria.

Usage

filter_candidate_motifs(
  find_candidate_motifs_results,
  sil_threshold = 0.5,
  size_threshold = 2,
  K = find_candidate_motifs_results$K,
  c = find_candidate_motifs_results$c
)

Arguments

find_candidate_motifs_results	Output from the find_candidate_motif function, which contains the results of the motif discovery process.
sil_threshold	A numeric value representing the threshold for the average silhouette index. Values should be between -1 and 1.
size_threshold	An integer representing the threshold for the size of the motif, defined as the number of curves in the cluster. Should be at least 1.
K	A vector containing the numbers of motifs that must be considered. This should be a subset of find_candidate_motifs_results$K.
c	A vector with the minimum motif lengths that must be considered. This should be a subset of find_candidate_motifs_results$c.

Details

This function checks the provided thresholds for silhouette indices and motif sizes, ensuring they are valid before filtering the motifs. If any parameters are invalid, default values from the find_candidate_motifs_results are used.

The function then loads the necessary data files for each combination of motif numbers and sizes, filtering them using the silhouette criteria. The resulting motifs are sorted by their length in descending order, and the filtered results are returned in a structured list.

Value

A list containing the filtered candidate motifs and some ProbKMA options:

V0_clean	A vector of candidate motifs after filtering.
V1_clean	A vector of derived candidate motifs after filtering.
D_clean	A matrix of distances of candidate motifs from the curves after filtering.
P_clean	A matrix of probabilities of membership of the candidate motifs after filtering.
Y0	The original input data for the first curve.
Y1	The original input data for the derivative of the curve.
diss	The dissimilarity matrix used in motif discovery.
alpha	The weight parameter for the Sobolev distance.
w	Weights for the dissimilarity index across different dimensions.
max_gap	Maximum allowable gap between curves.

funMoDisco documentation built on April 16, 2025, 1:10 a.m.