How to validate fuzzy c means?

kimagha_art71kimagha_art71 Member Posts: 1 Contributor I
edited December 2018 in Help

Hi,

I am using Fuzzy C-means operator for clustering ut I dont know how to validate it (with SSe or DavisBouldin or something else).

I would be honored if you could help me.

Answers

  • Thomas_OttThomas_Ott RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,761 Unicorn

    Take a look at the Segmentation Performance operators, they're located in Validation > Performance > Segmentation

  • marcin_blachnikmarcin_blachnik Member Posts: 61 Guru

    Hi

     

    A think you use Fuzzy C-means (FCM) from my extension - Information Selection, and in this case proposed by Thomas solution would work limited, because Fuzzy C-means do not return "CentroidClusterMode" (even it use centroids) instead it returns ClusterModel as output. In this case you can have several options

    1) you can use Log operator blouse all clustering methods implemented in this extension returns a special logValue which is called CostFunctionValue - in case of FCM it return the final value of the objective function

    2) you can not use Cluster Distance Performance but instead you can use Performance IS (Clustering) which takes as input dataset and cluster centers and returns performance equivalent to Cluster Distance Performance.

    You can use Performance IS (Clustering)  operator for any clustering method which returns centroids, then you just need to deliver to the "pro" input example set with cluster centers.

    3) Use other segmentation performance measures

     

     

    I attach example process which shows the first two solutions

     

    Best

     

    Marcin

Sign In or Register to comment.