I am interested in clustering validation and performance measures for example I like to run various experiments on text data using different clustering algorithm and compare the results based on for example Recall, percesion and cluster purity , I already know the categories ahead (supervised clustering )
Hi, then you could use the Operator Cluster2Prediction. It will match a clustername of an existing cluster attribute with a classname in an existing label attribute, so that the global matching is optimal. Then the usual performance operators might be used.
Answers
unfortunately we did not manage to create them until now.
If you have a special question about one parameter, feel free to ask
Greetings,
Sebastian
Thank you for your reply
I am interested in clustering validation and performance measures
for example
I like to run various experiments on text data using different clustering algorithm and compare the results based on for example
Recall, percesion and cluster purity , I already know the categories ahead (supervised clustering )
I have the CVS rapidminer
Thank you
then you could use the Operator Cluster2Prediction. It will match a clustername of an existing cluster attribute with a classname in an existing label attribute, so that the global matching is optimal.
Then the usual performance operators might be used.
Greetings,
Sebastian
i have the latest CVS
there is no such operator called cluster2prediction
Thank you
did you check out the developer branch "Zaniah"?
Greetings,
Sebastian