CalibratedBinaryClassificationMetrics.LogLoss Property

Definition

Gets the log-loss of the classifier.

public double LogLoss { get; }
member this.LogLoss : double
Public ReadOnly Property LogLoss As Double

Property Value

Remarks

The log-loss metric, is computed as follows:

LL = - (1/m) * sum( log(p[i]))

where m is the number of instances in the test set.

p[i] is the probability returned by the classifier if the instance belongs to class 1,

and 1 minus the probability returned by the classifier if the instance belongs to class 0.

Applies to