I have logs with the following information:

date-time username view action action_data

These logs are generated from a web-application which consists of several views where the users can perform a variety of actions which are logged (which elements they click, if they hover somewhere, switch views etc).

My desired output would be one or several clickstreams with some diverging paths which describe the most common behaviour (this will in turn be used as input for which sections of the site to test).

Since this will be used for testing, the generated clickstream doesn’t need to be definitive/straight but can include some diverging paths (40% of users click button A here and 60% press button B).

I’ve been looking at Markov chains and associative analysis for solving this but can’t land in something definitive about which methods to use.

  • 2
    $\begingroup$ Is a model really necessary? Why not simply count which paths or which transitions are the most common? $\endgroup$
    – Gala
    Jun 11, 2013 at 17:29

1 Answer 1


An alternative question about clicksteam data provided this interesting Microsoft paper about cluster clickstreams using first-order Markov chains. You may find it helpful.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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