Hi,
I have a bunch of measures from a Scorer node ("Accuracy statistics" output written to disc) that scores a binary classifier. I need to know how Knime determines the order of class 0 (Row ID = "0") and class 1 (Row ID = "1") in its output before the "Overall" row--> please see the attached image for a better understanding. I need to know this because I forgot to write the Row ID to disc and now I am not able to distinguish whether I need to read the result of row 1 or row 2 :(
Kind Regards
Hi Stev-io,
When looking in the "Accuracy statistics" output of Scorer node the RowId represents the class value.
In your example in the left Table: the first line with Row ID "0" stands for objects of class "0". In your case, the model predicted 23 objects as zeros, from which 19 entries were true zeros (True Positives) and 4 objects were non- zeros (False Positives) and so on for the following rows and columns. Therefore the "Accuracy Statistics" should not be read as a confusion matrix. To get the confusion matrix directly please check the output "Confusion Matrix" of Scorer node.
Hope it helps.
Best, Daria
Hi Daria,
thank you for your answer. But my problem is (was) a little different one. It is possible to reconstruct the confusion matrix from class 1 row in this binary case but the order of the rows output differs (if row index 1 stands for class 1 or class 0 seems to be selected whether the first member to learn is of class 1 or 0). I did not recognize this and evaluated only the row index instead of the row ID. I needed to change this and recalculated all the models.
Kind regards