I have been ding some exersizes with knime cluster analys and encontered with one issue.
I have 8 labeled types of disease and 20k features (float drype aka D) and 1000 samples(patient). I use k-means node. I recived 8 clusters but I don’t know what type of disease KNIME assign to each cluster.
Cluster_0 , 1, 4 , 5,…
Cluster_1 , 2, 4 , 7…
I try to use column appender after k-means node but I suppose it’s wrong because the score is unrealistic.
In other words I would like something what call in WEKA software “Classes for claster evaluation”
Which node distinctly assign label?
Cluster_0 to disease_5
Cluster_1 to disease_6
Thank you for your answers!
How is that working in WEKA? I would assume more diseases are matched to one cluster so how could you assign a distinct label to it?
I could only imagine mapping the disease column back (if same order column appender node could be used, then do a groupby cluster and concatenate / create a set of disease names for each cluster with the groupby node
Thank you for your comment!
I would explain what I means when quote Weka:
The problem is that I dont know which cluster Knime choose internaly. Also if I label (brain_type3 for Cluster_0 in Knime, the score is almost 0, because it is wrong association, but it is work for Weka
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.