Normal and anomalous value predefinition in case of unsupervised methods

Created by: igabriel85

Unsupervised methods usually consider 0 to be the normal label and 1 to be anomalous. However, some instances consider 1 to be normal and -1 to be anomalous.

Current solution is to autodetect this preference (both for training and prediction). Make it also user definable. See #6