K-means number of clusters

18 views
Skip to first unread message

Khaoula Ajbal

unread,
Nov 27, 2016, 8:48:33 AM11/27/16
to WekaMOOC-general
Hi, 

I am working on a dataset that I would like to cluster using k-means. But in the process I  had to choose the varial K, and since it affects the results i was wondering is there an optimal way to choose that variable? I read on forums about the elbow effect and the square root of the number of data points divided by two, but with about 500 instances, my k equaled 14 or so. and I couldn't conclude much out of it . 
Does weka provide a solution for this?

Best regards, 
Khaoula 

Ian Witten

unread,
Nov 28, 2016, 8:06:18 PM11/28/16
to wekamooc...@googlegroups.com
Yes. You should register for Data Mining with Weka and look at the Activity associated with Lesson 3.6 (https://weka.waikato.ac.nz/dataminingwithweka/activity?unit=3&lesson=6).
ian

--
You received this message because you are subscribed to the Google Groups "WekaMOOC-general" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wekamooc-gener...@googlegroups.com.
To post to this group, send email to wekamooc...@googlegroups.com.
Visit this group at https://groups.google.com/group/wekamooc-general.
To view this discussion on the web, visit https://groups.google.com/d/msgid/wekamooc-general/2d0a1d88-38c9-42a6-a743-6bbbc6bf9f1e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Khaoula Ajbal

unread,
Nov 29, 2016, 3:20:52 AM11/29/16
to wekamooc...@googlegroups.com

Thank you for your response.


Le 29 nov. 2016 00:06, "Ian Witten" <ian.w...@gmail.com> a écrit :
Yes. You should register for Data Mining with Weka and look at the Activity associated with Lesson 3.6 (https://weka.waikato.ac.nz/dataminingwithweka/activity?unit=3&lesson=6).
ian
On 28/11/2016, at 2:48 AM, Khaoula Ajbal <ajbal....@gmail.com> wrote:

Hi, 

I am working on a dataset that I would like to cluster using k-means. But in the process I  had to choose the varial K, and since it affects the results i was wondering is there an optimal way to choose that variable? I read on forums about the elbow effect and the square root of the number of data points divided by two, but with about 500 instances, my k equaled 14 or so. and I couldn't conclude much out of it . 
Does weka provide a solution for this?

Best regards, 
Khaoula 

--
You received this message because you are subscribed to the Google Groups "WekaMOOC-general" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wekamooc-general+unsubscribe@googlegroups.com.
To post to this group, send email to wekamooc-general@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "WekaMOOC-general" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wekamooc-general+unsubscribe@googlegroups.com.
To post to this group, send email to wekamooc-general@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages