0
$\begingroup$

we run some studies using google search queries.

We need to cluster these queries in topics and we would like to find some unsupervised approach.

for each query we have the search results.

I thought to use the number of urls in common as discriminants. Other discriminants could be the text similarity of queries and of titles of the results.

Since i'm quite new to clustering could you suggest me some approach?

Description and justification of the algorithm (and if possible implementation in R) is welcomed.

Thanks!

Improve this question
$\endgroup$
3
  • $\begingroup$ Maybe topic modeling is better suited than clustering? $\endgroup$
    – Has QUIT--Anony-Mousse
    Commented Sep 17, 2014 at 21:30
  • $\begingroup$ More info about it and the difference between the two? $\endgroup$
    – Bakaburg
    Commented Sep 18, 2014 at 11:08
  • $\begingroup$ Google for it. Topics is designed for text, while most clustering methods are designed for dense "point" vectors with Euclidean distance. $\endgroup$
    – Has QUIT--Anony-Mousse
    Commented Sep 18, 2014 at 11:28

0

Reset to default

Your Answer

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

Browse other questions tagged or ask your own question.