what is cluster Analysis?

and how to do it?

[seoram]

Can anyone tellme about cluster analysis and how is that helpful in knowing the site analytics. Also most important how can we do that analysis?

[cgrantski]

If you Google "cluster analysis" you will find a lot of information on it. It's a way of discovering how users view pages or site sections in groups. It's also a way of discovering groups of users who have the same general site behavior. It is really, really, REALLY valuable. Far more than the tabulations that analytics packages do, IMHO.

None of the analytics packages commonly mentioned here can do cluster analysis, that I know of. SAS and SPSS are the main programs I know of that can do it for web sites. To get into this kind of thing you need many thousands $$$.

[cgrantski]

To be more complete about my answer --- I believe Visual Sciences has a type of cluster analysis available.

And, for certain ordinary stats packages (WebTrends and Omniture for sure, not definite about others) it is possible for a statistically knowledgeable person to take certain output from those packages and, with help from Excel and a cheap stats package, do a site-content type of cluster analysis with a fair amount of work. The visitor type of cluster analysis would then be a few more labor-intensive steps.

[McElvoy]

I'd like to know where you saw references to cluster analysis and web analytics. They are not often associated.

[alexdino1]

I use both Visual Sciences as well as SAS and can say that both are capable of many different kinds of analysis, but Visual Sciences technically does not do Cluster analysis. They do an excellent job with segmentations, at both visitor and session levels.

Cluster analysis is finding groups of visitors based on specific attributes. For example, by using SAS I can find a cluster of visitors based on similar page views per session distributions. SAS will dictate the clusters.

Segmentation analysis, on the other hand, takes a more simple approach. You have to manually set up your segment. For example Segment A are visitors with less than two page views per session, segment B are visitors with between two and five page views per session, and segment C are visitors with greater than 5 page views per session.

[seoram]

Thanks a ton for all the knowledge that you have given me.
I found that when i was reading a blog

[hwjgage]

Hi, you can extract the data from your web analytics tool and do cluster analysis in excel - I find using 3D bubble charts is quite a nice way to display the data. In order to get decent data from you web analytics tool you will need to make sure your content naming convention is up to scratch and may also want to emply segmentation. Mmost decent web analytics tool will give you segmentation although Google Analytics does it in a rather half arsed way.

Hope this helps.

[cgrantski]

hwjgage, cluster analysis has to do with co-occurrence of things within the same event (visit or visitor, in this case). How are you getting your stats program to output the % of time any two particular pages happen within the same visit?

Or are you referring to a scatter plot? What are the axes and bubble size variable in your graph? If you are managing to do a real cluster analysis I'd love to know how.