Table
1: Metrics.
Metric |
Description |
Formula |
TPR (sensitivity
or recall) |
True
Positive Rate ·
probability
of a positive test, given a patient ill; ·
measure
of prediction correctness; ·
should
be high. |
|
PPV (or
precision) |
Positive
Predictive Value ·
the
proportion of the true positives against all the positive results; ·
capacity
of predicting the positives; ·
used
instead of FPR; ·
should
be high. |
|
F-measure |
F-measure ·
harmonic
weighted mean between PPV and TPR; ·
close
to 1 = better prediction. |
|
MCC |
Matthew’s Correlation Coefficient ·
equal
to: - -1: false assignments (TN = TP = 0) - 0: prediction
not better than random - 1: all true
assignments (FP = FN = 0) Set to 0
when denominator = 0 ·
~
geometric mean between TPR and PPV; ·
should
be high. |
|
TP = True Positive; TN = True Negative; FN = False Negativ;
FP = False Positive