Building the confusion matrix

Let's now think about a binary classification problem. We have a set of samples belonging to two classes: YES or NO. We can build a machine learning model that outputs a class for each input set of variables. By testing our model on 200 samples, we will get the following results:

There are four elements to the confusion matrix:

• True positives (TP): The number of times that the model predicts YES and the actual value is YES. In our example, this is 100 times.
• True negatives (TN): The number of times that the model predicts NO and the actual value is NO. In our example, this is 60 times.
• False positives (FP): The number of times that the model predicts YES and the actual value is NO. In our example, this is 15 times.
• False negatives (FN): The number of times that the model predicts NO and the actual value is YES. In this example, this is 25 times.

Then, we calculate the confusion matrix in the following equation: