DoTC (version 0.2)

wrapFKM: Wrapper for FKM

Description

Wrapper function for a call to fuzzy k-means function FKM.

Usage

wrapFKM(d, m, start, maxit = 1e4, threshold = 0.1)

Arguments

d

data-set with standardized columns

m

Fuzziness parameter

start

Starting values as provided by getStart.

maxit

Maximum number of iterations (default is 10000).

threshold

Upper limit below which a cluster distance (as defined by matrix H in FKM) is set to be 0.

Value

FKM:U (matrix containing case-wise (rows) cluster (columns) affiliation values), H (pair-wise cluster distance matrix), value (terminal value of the fuzzy-clustering algorithm), iter (number of iterations needed to get to value), k (number od proposed clusters to the start solution start), call (call to the interior FKm function), combined (which clusters are members of any combinations), remaining (which clusters stay remaining), combinations (which are the combinations that lead to the reduction), and n_cluster (number of reduced clusters),and further attributes of the solution across potential different proposal fuzziness values:m (all proposed fuzziness values), n_cluster (the number of reduced clusters), m_before_step (the maximum fuzziness parameter before a reduction in n_cluster), and which_list_indexes_m_before_step (where are the respective results to m_before_step).

Details

The function is a specific wrapper function to a function which gives the same results as FKM from the R package fclust.

The below example gives an exemplary complete run for an analysis as implemented by this package DoTC.

The selection of the fuzziness parameter m is crucial for the result of the wrapFKM function. We have good experiences with following strategy: first use a coarse grid of proposal values for m, look on aggregation of clusters across this proposal vector -- as for example implemented by getStart --, and then refine this grid for a certain sub-interval of interest.

References

Paolo Giordani, Maria Brigida Ferraro (2015). fclust: Fuzzy Clustering, on CRAN.

Examples

## Not run: 
# ## Load and standardize (by column) data:
# d <- read.csv("data_file.csv")
# d <- apply(d, MAR = 2, FUN = scale)
# ## Set maximal number of clusters:
# K <- 10
# ## Set random seed:
# set.seed(1604)
# ## Get k-means-clustering solutions as starting values:
# start <- getStart(d = d, K = K)
# ## Proposal vector for fuzziness parameter m:
# m_proposal <- seq(1.1, 2.5, by = 0.1)
# ## Calculate results of fuzzy clustering:
# fkm_result <- wrapFKM(d = d, m = m_proposal, start = start)
# ## Plot cluster solution across varying m:
# plotNcluster(fkm = fkm_result)
# ## Plot distribution of typicality coefficients:
# plotTC(fkm_result[[1]])
# ## Plot pairwise cluster segregation comparisons:
# plotCS(fkm_result[[1]])## End(Not run)