If different types of errors are associated with different costs a weighted error function might be more appropriate than the standard.
Cost matrix or factor response vector.
This function is not in itself an error function, but used to generate error functions. Either supply a predefined cost matrix or a response vector for a classification problem to define it automatically.
The automatically generated cost matrix will generate an error of 0 if all predictions are correct, 1 if all predictions are incorrect and 0.5 if all predictions are the same (regardless of class, i.e. if one class is smaller it will be given a higher misclassification cost).
An error function.