I want to perform cluster analysis on the data of a website. The data is mainly visitor history(97000 rows) and has following variables:

a)User Device Category b) Traffic Marketing Channel c) Traffic Source d) Marketing Campaign e) Website Landing Page f) User Session Duration g) User Goal Completion

Apart from session duration and goal completion rest of the data is categorical. In particular traffic source,campaign,landing page etc have 100's of levels each.

Also I believe the goal completion is sparse as most of the time it is 0.

Would it be a good idea to:

  1. bin the session duration to categories. eg. 1000 seconds to 3000 sec.is one category and so on

  2. if a level of categorical variable is not too frequent then just write "others"

  3. combine certain levels if they have similar characteristic.

Could someone give me tips on how to handle the categorical data, which package in R can handle mixed data type and what would be a good way to interpret the results.

I am fairly new to data analysis and need all the help I can get. Thanks a lot in advance!

  • 1
    $\begingroup$ Please try first to read posts containing some words from clustering + categorical mixed sparse. That theme is an old one on the site. $\endgroup$ – ttnphns Jun 4 '16 at 9:14

I don't think cluster analysis is the proper tool here.

By any means, the different categories simply are different categories. There is no reason to mix them up!

What you want to do is aggregation and summary statistics, as available in the standard toolkits such as G__gle Analytics and Piwik. There is a reason why G__gle Analytics does not "cluster" the visits (and it's not that G__gle doesn't know how to cluster) but provides complex toold duch as funnel analysis instead.


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