I've read node description for ROC Curve and really confused. How this description connected with other definitions (with specifity and sensivity)? What are variables on axies shown?
Maybe there is any article, where I can read about this?
Wikipedia has a nice article about ROC curves.
Thank you, thor,
Wikipedia and more significant pages for me (I'm russian, so page on Russian language, sorry, but it has algorithm) uses algorithm with sensivity and specifity calculation. Here, on KNIME site, I see another version of algorithm, so my question is in equivalency of this two versions and proofs of it.
KNIME implements the standard ROC procedure: the y-axis shows the true positive rate, the x-axis the false positive rate, which are equal to sensitivity and (one minus the) specificity.
I reopen this topic because unfortunately with my data I don't see how to use this node and the description does not help me.
That would be great if you could provide a workflow example. At least, I would see how it works and the input data.
We don't have an example workflow for ROC yet but the node description is fairly detailed... Probably you can describe your use case and we can create an example from it?
Yes, you are right, after several trials, I finally understood what you mean in the description.
Bro can you share the workflow?
i don't know how to plot any ROC curve in KNIME, and i want to plot multiple ROC curves in one line chart.
Please Guide me.
There is a dedicated ROC Curves node. For comparing multiple classifiers in one chart, join the classification results together using the Joiner node and select the classification probability columns to plot in the ROC Curves node settings.