Skip to content

An R package to improve workflow with k-modes models.

License

Unknown, MIT licenses found

Licenses found

Unknown
LICENSE
MIT
LICENSE.md
Notifications You must be signed in to change notification settings

Sdbock/kmodesFlow

Repository files navigation

kmodesFlow

The kmodesFlow package provides convenience functions to improve workflow with kmodes modeling. It allows users to easily fit multiple models, assess model fit, and examine results with automated visualizations. See full demo here.

Installation

You can install the development version of kmodesFlow from GitHub with:

# install.packages("devtools")
devtools::install_github("Sdbock/kmodesFlow")

Tutorial

The workhorse of the kmodesFlow package is the fit_models() function. To fit 8 models, with 1 through 8-k solutions, simply specify “1:8” as the number for k:

library(kmodesFlow)

fit_models(
  data = data,
  k = 1:8,
  seed = 1234
)

The fit_models() returns a tibble with the number of rows equal to the number of models specified. The columns contain information about cluster assignment for respondents, model fit, and visualization of cluster profiles. To examine the modal category for each variable within clusters, print the table_cluster_modes column for the relevant model. I suggest using the pluck() function to identify and print the desired row and column. To look at the modes across clusters for the k = 3 model:

pluck(model_output, "table_cluster_modes", 3)

There is also a heat map of the distributions of variable levels within each cluster, which can be accessed in the “table_attribute_distribution” column:

pluck(model_output, "table_attribute_distribution", 3)

Finally, the plot_elbow() function takes a fit_models() output and returns a plot of the sum of within-cluster differences across each level of k. This “elbow” plot is a common tool in determining the optimal number of clusters.

plot_elbow(model_output)

About

An R package to improve workflow with k-modes models.

Resources

License

Unknown, MIT licenses found

Licenses found

Unknown
LICENSE
MIT
LICENSE.md

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published