Methods to determine best number of clusters

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Methods to determine best number of clusters

haytham.salhi
Hello Weka, 

While searching for methods in Weka to determine the best number of clusters (K) for k-means algorithm, I found CascadeKMeans.

However, there are indeed other methods like elbow method for that purpose. My question is two fold: 

Are there currently any other implementation in Weka to do so? 

Is it possible to implement elbow method to find the best k without plotting the elbow (i.e., find best k programmatically based on elbow method)?

Best,
Haytham

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Methods to determine best number of clusters

Eibe Frank-2
Administrator

> On 31/05/2017, at 2:43 PM, Haytham Salhi <[hidden email]> wrote:
>
> While searching for methods in Weka to determine the best number of clusters (K) for k-means algorithm, I found CascadeKMeans.
>
> However, there are indeed other methods like elbow method for that purpose. My question is two fold:
>
> Are there currently any other implementation in Weka to do so?

There is also XMeans.

> Is it possible to implement elbow method to find the best k without plotting the elbow (i.e., find best k programmatically based on elbow method)?

Yes, if you have a suitable mathematical criterion for detection the location of the elbow, you can write a Java/Jython/Groovy script that runs SimpleKMeans multiple times for different K and grabs the sum of squared errors using the

  getSquaredError()

method.

Cheers,
Eibe

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Loading...