How do we perform cross cluster analysis in R. Most of my data has categorical variables {variable Marital Status (married, single, divorced); variable Education (tertiary, secondary, primary, etc.); and the variables Credit, Mortgage, and Loan (yes, no)}? The only interval variable in my data set is the Age.

  • $\begingroup$ Questions that are only about how to use software are generally off topic here. If you have a machine learning question about clustering categorical data, please edit to clarify. $\endgroup$ – gung - Reinstate Monica Sep 27 '17 at 14:33
  • 2
    $\begingroup$ I think this person is very early in learning. I don't think they are asking R only. I think they are asking "how do I think about this". It is exceptionally broad, but I think they are asking for "starting points" to enter the subject. I don't think they know the questions they could ask. $\endgroup$ – EngrStudent Sep 27 '17 at 14:38

Possible duplicate here, here and here. Distance-based clustering algorithms can handle categorical data. So you can implement clustering from a dissimilarity matrix.

First, you have to compute all the pairwise dissimilarities (distances) between observations in the data set (with daisy()). Then, you can run your clustering algorithm (with agnes(), CrossClustering(),...).

Here is an example.

flower <- flower[, 1:6] # just to keep only categorial variables
# Dissimilarity matrix matrix
distm <- daisy(flower, metric = "gower", stand = FALSE)

# Hierarchical agglomerative clustering (HAC)
hac <- agnes(distm, diss = TRUE)

# Partial clustering algorithm with automatic estimation of the number of 
# clusters and identification of outliers
cross.clust <- CrossClustering(distm, k.w.min = 2, k.w.max = 5,
                               k.c.max = 6, out = TRUE)
  • $\begingroup$ Thank you ANG. That was really helpful. And sorry if my question looks similar to others. $\endgroup$ – Lav Sep 28 '17 at 1:21
  • $\begingroup$ Error in CrossClustering(distm, k.w.min = 2, k.w.max = 5, k.c.max = 6, : could not find function "CrossClustering" I tried to correct last two lines. Is this correct code? cross.clust <- cc_crossclustering(distm, k_w_min = 2, k_w_max = 5, k2_max = 6, out = TRUE) cross.clust $\endgroup$ – vasili111 Oct 18 '19 at 23:46

Not the answer you're looking for? Browse other questions tagged or ask your own question.