2
$\begingroup$

I'm a SQLServer DBA and in the new version of this tool, there are new features to integrate R scripts and use it easily with the DB objects. That sounds cool.

But to use that, we have to know a little bit about stats and data mining. And it's quite hard to find friendly and readable documentation for newbies like me. So, I will ask for some best-practice about clustering.

I have a dataset with few hundreds records and with 5 features, both categorical and numerical.

Regarding categorical data, I have a few basic questions :

  • how can I manage categorical data ? what I understood from articles found is that I can transform categorical data in boolean and "pivoting" it. If a have a column with categorical data (for instance with values 'a' or 'b'), creating to new columns like column_a and column_b with value 0 or 1 ?

  • should I always center / scale this new column ?

$\endgroup$
  • $\begingroup$ Welcome to the site. I think the question can be broken down into multiple questions. Points 1&2 together, 3, 4, 5 together and the last one separately. :) $\endgroup$ – Dawny33 Apr 27 '16 at 12:50
  • $\begingroup$ Indeed, let's just keep the categorical topic. (title and question have been rephrased) $\endgroup$ – irimias Apr 28 '16 at 11:37
1
$\begingroup$

You need to create a dissimilarity matrix first and then apply the clustering technique. Below link has many answer referring your point. Hope below will suffice for this:

https://stats.stackexchange.com/questions/24540/clustering-of-mixed-type-data-with-r

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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