Dynamic D3.js Based K-Means Clustering Visualizations in R

This package provides methods for dynamically visualizing k-means clustering data or any ordinal data and its associated clusters, though the original intention was to provide users with a more user friendly visualization tool for k-means clustering.

Development Version: 0.1.0.9000 Travis build status CRAN status

Installation

Use requires package htmlwidgets.

library(devtools)
install_github("ramnathv/htmlwidgets")
install_github("McKayMDavis/klustR")

Basic Usage

pcplot, a dynamic visualization of dimensionally reduced data:

scaled_df <- scale(state.x77)
clus <- kmeans(data_scaled, 5)$cluster
pcplot(data = data_scaled, clusters = clus)

Things to note:

  • Clicking on an axis label will display a bar-chart of each column’s contribution percentage to that particular dimension or principal component

  • Hovering over points displays the label

  • Clicking on a color on the legend highlights the associated cluster

pacoplot, a dynamic parallel coordinates plot:

df <- state.x77
clus <- kmeans(data_scaled, 5)$cluster
pacoplot(data = df, clusters = clus)

Things to note:

  • Hovering over a line displays the label

  • Clicking on a line highlights the associated cluster

  • Clicking on the “Toggle Averages” box displays median lines and 1st and 3rd quartile intervals for each cluster

Visit website for more details.