In relation to web usage mining from a log file, can you cluster data without performing User and/or Session identification?

I mean,let's say I have these entries:

123.234.324.122 [timestamp] "GET /cars/sport/porsche.jpg" 200 23432 "http://topgear.com/cars" "Mozilladsfsd"

120.23.324.122 [timestamp] "GET /bikes/sport/r1.jpg" 200 23432 "http://topgear.com/cars" "Mozilladsfsd"

13.234.324.122 [timestamp] "GET /cars/utility/micra.jpg" 200 23432"http://topgear.com/cars" "Mozilladsfsd"

So,in this scenario, I just need to cluster based on which cars have been viewed more frequently etc etc..Do I need user identification and session identification then? Or can I just consider the URLs and cluster on them?

Because as far as the traditional Web Usage Mining approach goes and all the papers I've gone through suggest, you do preprocessing,then the pattern-discovery comes along..

My question is why not jump to the pattern discovery straight-away????

  • $\begingroup$ You may want to be a bit more specific about your data and what you want to do with it? $\endgroup$
    – crayola
    Commented May 11, 2011 at 14:30
  • $\begingroup$ Specificity provided :) $\endgroup$
    – lightsong
    Commented May 11, 2011 at 14:54

2 Answers 2


If you don't identify your sessions/users, you are clustering different things: one user who is an insane adept of any given car and looks at its picture dayly could have a huge impact on your results, though you're probably not interested in this.


Cluster analysis does not involve hypothesis testing per se, but is really just a collection of different similarity algorithms for exploratory analysis. You can force hypothesis testing somewhat but the results are often inconsistent, since cluster changes are very sensitive to changes in parameters. So the answer is yes, you can do it, but be careful about making specific statistical inference.

As to your other point. As in any data analysis, your data should be as clean and as representative as possible, so I would avoid jumping steps.


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.