Metrics for Classifiers

In the previous section, you learned how to train a binary classifier to predict the right output: either 0 or 1. In Exercise 5.01, Building a Logistic Regression Model, you looked at a few samples to assess the performance of the models that were built. Usually, you would evaluate a model not just on a small subset but on the whole dataset using a performance metric such as accuracy or F1 score.

Accuracy and Null Accuracy

One of the most widely used metrics for classification problems is accuracy. Its formula is quite simple:

Figure 5.14: Formula of the accuracy metric

The maximum value for accuracy is 1, which means the model correctly predicts 100% of the cases. Its minimum value is 0, where the model can't predict any case correctly.

For a binary classifier, the number of correct predictions is the number of observations with a value of 0 or 1 as the correctly predicted value:

Figure 5.15...