Unsupervised clustering of categorical data

Hi everyone,

I wanted to cluster a time serie dataset which has 30 timepoints and more than 50'000 rows. The dataset is categorical (from 1 to 6) which represent different categories.

The problem with my current clustergram method using the euclidian distance metrics, is that it will cluster the category 5 closer to 6. I don't want that, those categories are not somehow related. How is it possible to remove this bias in the clustering?

Hope my question is clear, thanks for your further help!

[Image Analyst]

Could be clearer if you attached a .mat file with your table, as many rows as will fit into 5 MB (attachment size limit).

[Daniel Guignard]

sure, here is the matrix

Answer 1

[Pratyush Roy]

Hi Daniel,

The link here might be helpful for clustering categorical or non-numeric data.

Hope this helps!