7.3 Positive and Negative Outcome Classes
In order to define the next several metrics, we will start with some terminology.
In two-class scenarios, we label one class as “positive” and the other as “negative.” When thinking about these terms, we need to dissociate them from the way we think of these words. There is nothing necessarily “good” about the positive class, and nothing necessarily “bad” about the negative class. The designation of one class as “positive” and the other as “negative” is purely arbitrary.
When modelers talk about prediction outcomes, they often talk about terms like True Positive, False Positive, True Negative and False Negative. Sometimes those terms can be confusing! Sometimes, it may be easier to think about them by inserting the word ‘Predicted’ in the middle of the terms.
A True Positive, or a True Predicted Positive, occurs when the model predicts that the record will belong to the positive outcome class, and that prediction turns out to be correct.
A False Positive, or a False Predicted Positive, occurs when the model predicts that the record will belong to the positive outcome class, and that prediction turns out to be wrong.
A True Negative, or a True Predicted Negative, occurs when the model predicts that the record will belong to the negative outcome class, and that prediction turns out to be correct.
A False Negative, or a False Predicted Negative, occurs when the model predicts that the record will belong to the negative outcome class, and that prediction turns out to be incorrect.
For each of the examples below, we will use “renew” as the positive outcome class, and “non-renew” as the negative outcome class.
| actual renew | actual non-renew | TOTAL | |
| predict renew | TP | FP | TP+FP |
| predict non-renew | FN | TN | FN + TN |
| TOTAL | TP + FN | FP + TN | TP + FP + FN + TN |