leaderSO: Adapted leaders clustering of modal multi valued symbolic...

In clamix: Clustering modal multi valued symbolic data

Description

Function to compute adapted leaders clustering for modal multi valued symbolic data. Data must be saved in a symData object.

Usage

leaderSO(dataset, maxL, initial = NULL, stabil = 1e-06, report = TRUE, interact = TRUE,type = "d1")

Arguments

dataset	symData object
maxL	number of clusters to produce
initial	provide initial leaders for clustering; if NULL they are selected randomly
stabil	stability of the results (how close should be the minimization functions of two algorithm steps to exit)
report	produce report while working
interact	do you want interaction during algorithm steps
type	switch describing dissimilarity measure used "d1" to "d6" or "d1w" (calculating with frequency distributions "fDist") (for measures see Table 1 in reference article below)

Value

Returned is a list of elements: list(clust=clust,leaders=L,R=R,p=p)

clust	partition of clusterings
leaders_symData	symData object of leaders for each cluster (for possible direct input into hclustSO function)
R	minimal error (distance to leader) for each cluster
p	sum of errors for each cluster

Author(s)

Vladimir Batagelj

References

V. Batagelj, N. Kejzar, and S. Korenjak-Cerne. Clustering of Modal Valued Symbolic Data. ArXiv e-prints, 1507.06683, July 2015.

See Also

hclustSO, create.symData,print.symObject

Examples

data(popul06f)
data(popul06m)
datalist <- list("M"=popul06f,"F"=popul06m)
dataset <- create.symData(datalist,"pDist")
# type = "d1w" can be used for dataset created with "fDist"
## Not run: 
res2006 <- leaderSO(dataset,5)
# Times repeat (if <=0, no repetitions)
summary(res2006$leaders_symData)
print(res2006$leaders_symData$SOs)

## End(Not run)

clamix documentation built on May 2, 2019, 4:51 p.m.