Active cost-sensitive learning

摘要: For many classification tasks a large number of instances available for training are unlabeled and the cost associated with labeling process varies over input space. Meanwhile, virtually all these problems require classifiers that minimize nonuniform loss function decisions (rather than accuracy or errors). example, to train pattern models network intrusion detection task, experts need analyze events assign them labels. This can be very costly procedure if labeled selected at random. In meantime, mislabeling an is much higher opposite error (i.e., legal event as being intrusion). As result, address types tasks, practitioners tools total computed sum decisions. This paper describes approach addressing this problem.