Instillation and running of the pandemonium app was done with this code.
devtools::install_github("uschiLaa/pandemonium")
library(pandemonium)
pandemonium(b_anomaly$pred, b_anomaly$covInv, b_anomaly$wc, b_anomaly$exp)
Within the app the following settings used to compare their clustering results:
-
A. Linkage - complete, include covariance - true, coordinates - pull, distance - euclidean, clusters - 5
-
B. Linkage - complete, include covariance - false, coordinates - pval, distance - manhattan, clusters - 2
From these setting it was found that B split the clusters based on the value of c9 whereas A had much more interesting cluster assignment.
A grand tour and a local tour were shown side by side using detourr R package in a shiny app. The iris data set was used. crosstalk was used for linked brushing.
visuals can be found here.