Skip to contents

Estimation of the best number of clusters using the Kamila algorithm

Source: R/mod_kstar.R
mod_kstar.Rd

Estimation of best number of clusters using the Kamila algorithm on the training set.

Usage

mod_kstar(PARAM_KAMILA, CATEG_DF_TS, CONT_DF_TS, list = NULL)

Arguments

PARAM_KAMILA

data frame with all needed parameters for the Kamila method, from which the following parameters are used:

  • numberofclusters: The number of clusters returned by the algorithm, i.e. sequence indicating the number of clusters which should be investigated to extract the optimal number of clusters.

  • numinit: The number of initializations used.

  • maxiter: The maximum number of iterations in each run.

  • calcnumclust: Character: Method for selecting the number of clusters. Setting calcNumClust to ’ps’ uses the prediction strength method of Tibshirani &Walther (J. of Comp. and Graphical Stats. 14(3), 2005).

  • pred_threshold: Threshold fixed to 0.8 for well separated clusters, i.e. not overlapping.

CATEG_DF_TS

Training set of the pension register containing all categorical variables as factors.

CONT_DF_TS

Training set of the pension register containing all the continuous variables.

list

List of input data frames.

Value

a tidylist containing the following tidy data frames:

  • KM_RES Data frame containing the results of the clustering.

  • PARAM_KAMILA Data frame with the updated kstar parameter.

Author

Layal Christine Lettry