fc_rclust: Generate a List of Random kcca Objects.

In ds4ci/CustSegs: Tools for Customer Segmentation

Description

For a given number of clusters, k, nrep kcca objects are generated.

Usage

fc_rclust(x, k, fc_cont, nrep = 100, fc_family, verbose = FALSE,
  FUN = kcca, seed = 1234, plotme = TRUE)

Arguments

x	Integer. matrix. Input to kcca.
k	Integer. Number of clusters for this run.
nrep	Integer. Number of repititions to run.
verbose	Logical. Override for fc_control slot verbose.
FUN	flexclust function.
seed	Integer. Starting set.seed value for this run.
plotme	Logical. Should plot be produced as side-effect?
fc_contol	The flexclust control object for this run.

Details

Each is kcca object is re-ordered so clusters are in decending size order. Cluster summary information is pulled out of each object from the clusinfo slot. This includes the sizes of the clusters.

To look for stable cluster solutions, each random run is characterized by the sizes of the first two clusters; which will be the largest after reording. kde2d() from the MASS package is used to find density contours. The highest peak is determined and the distance of each solution to the peak is retained for each cluster.

Optionally, the scatter plot of the sizes and corresponding coutour in plotted.

Value

A list(best, sizes, peak_at, tries)

ds4ci/CustSegs documentation built on May 15, 2019, 2:56 p.m.