Machine Learning With Go

Categorical metrics

Let's say that we have a model that is supposed to predict some discrete value, such as fraud/not fraud, standing/sitting/walking, approved/not approved, and so on. Our data might look something like the following:

observed,predicted
0,0
0,1
2,2
1,1
1,1
0,0
2,0
0,0...

The observed values could take any one of a finite number of values (in this case 1, 2, or 3). Each of these values represents one of the discrete categories in our data (class 1 might correspond to a fraudulent transaction, class 2 might correspond to a transaction that is not fraudulent, and class 3 might correspond to an invalid transaction, for example). The predicted values could also take one of these discrete values. In evaluating our predictions, ...