Table 2.
Actually positive | Actually negative | |
---|---|---|
Predict positive | Cost(+, +) | Cost(+, −) |
Predict negative | Cost(−, +) | Cost(−, −) |
Cost values are set according to different classification results. Generally, for a rare positive and prevalent negative samples, Cost(+, −) > Cost(−, +). And Cost(+, +) = Cost(−, −) = 0 denotes no penalty for a correctly predicted sample.