Binned Performance


Calculates various performance measures (detailed below) on the input table. This node also takes a column to bin the values on, for example bin on a probability column into 5 bins of: 0.5 ... 0.6, 0.6 ... 0.7, 0.7 ... 0.8, 0.8 ... 0.9 and 0.9 ... 1.0. The min value of the first bin is inclusive and the min value is exclusive for all other bins. All ranges are inclusive of the max value.


Balanced accuracy: Sensitivity + Specificity / 2

Accuracy: TP + TN / 2

Sensitivity: TP / (TP + FN)

Specificity: TN / (TN + FP)

Precision aka Positive Predictivity (PPV): TP / (TP + FP)

Negative predictivity (NPV):TN / (TN + FN)

Recall: TP / (TP + FN)

F-Measure 2 * ((precision * recall) / (precision + recall))

Also outputs the counts for TP, FP, TN, FN, number of equivocals and number of out of domains and coverage (% not out of domain).

Note that the number of equivocals and number out of domain do not impact on the Cooper statistics (Sensitivity, specificity etc.)

Target values that do not match the active or inactive value specified are not included in the calculation.

Input Ports

  1. Type: Data Table containing at least the binning column, activity and prediction.

Output Ports

  1. Type: Data Row containing all the calculated measures per bin
  2. Type: Data A row may fail because of a missing value or a value outside the binning range. These rows are output in this table.

Find here

Community Nodes > Lhasa Limited > Generic > Scoring

Make sure to have this extension installed: