is_valid_cluster: Determine whether a cluster is valid

In zatonovo/kingsmen: A technique for detecting regime change in irregular time series

Description

If a cluster minimizes distance between points, then it is considered to be a valid cluster.

Arguments

z	An n x 2 matrix of points
groups	A vector of group associations

Value

A logical value is returned indicating whether the specified groups are considered valid.

Usage

is_valid_cluster is_valid_cluster(z, groups)

Details

With hierarchical clustering techniques it is difficult to know whether the set of clusters produced are valid. Typically this is left to interpretation and must be 'eye-balled' to choose the cutoff point as well as decide whether the cluster boundaries make sense.

For an automated system, such a manual decision point is undesirable and must be replaced by automatic process. Since this data is multidimensional, one approach is to use a distance metric or other mathematical property as a heuristic. This function uses accepts a group of clusters if the sum of the variances of the distance within each cluster is less than the variance of the distance as a single cluster.

Author(s)

Brian Lee Yung Rowe

zatonovo/kingsmen documentation built on May 4, 2019, 9:11 p.m.