Setup GMMConfig

Setup a GMMConfig object for Gaussian Mixture Model clustering, via the 'mclust' package.

Usage

setup_GMM(k = NULL, model_names = NULL, features = NULL)

Arguments

k

Optional Integer [1, Inf): Number of mixture components. Unset selects it by BIC.

model_names

Optional Character {"EII", "VII", "EEI", "VEI", "EVI", "VVI", "EEE", "VEE", "EVE", "VVE", "EEV", "VEV", "EVV", "VVV"} vector: Covariance parameterizations to consider. Unset considers all of them.

features

Optional Character vector: Names of at least 2 distinct feature columns to cluster on. NULL clusters on all numeric features.

Value

GMMConfig object.

Details

A GMM is the model-based counterpart of the prototype methods: each cluster is a Gaussian component with its own mean and covariance, so clusters may be elongated, differently oriented and differently sized, and every case gets a posterior probability for each component as well as a label. The result is therefore a soft clustering.

Both the number of components and the covariance parameterization are selected by BIC, which makes k optional: leave it unset to have the number of components chosen, or set it to fix the number and let BIC choose only the parameterization.

Argument names are rtemis' own: k is mclust::Mclust()'s G and model_names is its modelNames.

References

Scrucca L, Fop M, Murphy TB, Raftery AE (2016). mclust 5: Clustering, Classification and Density Estimation Using Gaussian Finite Mixture Models. The R Journal, 8(1), 289-317. doi:10.32614/RJ-2016-021

Author

EDG

Examples

gmm_config <- setup_GMM(k = 3L)
gmm_config

#> <GMM ClusteringConfig>
#>           k: <int> 3
#> model_names: <NUL> NULL