Fuzzy clustering operator in rapidminer?

ahootanha

Hello to all. Thanks to the good rapidshare community. I want to do fuzzy clustering (soft). Like c mean fuzzy. But I did not see the Operator in RapidMiner. Someone guides me?

And how to compare the clustering with the purity and NMI evaluation measure? What is their operator? Many thanks

ahootanha

Hello
I found a fuzzy cluster operator
And
I did clustering
How do I assess my clustering and compare kmeans with clustering?
??
How use nmi and purity measure ?

marcin_blachnik

See Information Selection Extension - it contains FCM. On the "ori" output port it returns clustered data and the partition matrix. "pro" output port contains cluster centers.

 

Take a look at the log value of the fcm operator. It contains the value of the cost function.

 

The "mod" output is ClusterModel which you can use to cluster new data.

 

To compare performances of different methods please use: Item Distribution Performance, Cluster Density Performance, or Performance IS (Clustering).

 

The last operator takes example set and cluster centers as inputs and then calculates the performance=sum(min(D(x,p)^2), where the sum is over every sample from the exampleSet, and the min - minimum distance is over all cluster centers. In other words it finds the nearest cluster center, and sum the squared distances.

 

 

---

ahootanha

Hello. Thank you very much for your kindness. I did, I sent the result. I wanted to ask you to guide me. How to set the parameter to give a better result? What does this answer result from the performance operator? Has the result been good? Sorry I do not ask to ask these questions but I am a beginner in fuzzy clustering. Thank you so much