User Tools

Site Tools


Performance curves


Brief description

Calculation of predictive performance measures like precision-recall or ROC curves.


Indicated are

Specification whether rows containing the “Indicator” in the categorical column specified in “In column” correspond to the class under observation or not (default: False).

In column

Selected categorical column containing the class membership of each instance (row) of the class under observation (default: first categorical column in the matrix).


Rows containing the defined string are counted as true or false depending on the selection in “Indicated are” (default: +).


Selected expression columns containing the scores by which the rows are ranked to calculate the specified quantities (default: first expression column of the matrix is selected).

Large values are good

If checked the larger the score value the better (default: checked). Otherwise the lower the value the better.

Display quantity

Selected quantities that will be calculated (default: no quantities are selected). The quantities can be selected from a predefined list:

  • TP/(TP+FP) (Precision)
  • TP/(TP+FN) (Recall)
  • FP/TP
  • TP/NP
  • TP/(TP+FN) (Sensitivity)
  • TN/(TN+FP) (Specificity)

Parameter window

Perseus pop-up window: Basic -> Performance curve

perseus/user/activities/matrixprocessing/basic/performancecurves.txt · Last modified: 2015/11/23 14:47 (external edit)