Machine Learning for OpenCV 4 - Second Edition

Measuring model performance with scoring functions

One of the most important parts of building a machine learning system is to find a way to measure the quality of model predictions. In real-life scenarios, a model will rarely get everything right. From earlier chapters, we know that we are supposed to use data from the test set to evaluate our model. But how exactly does that work?

The short, but not very helpful, answer is that it depends on the model. People have come up with all sorts of scoring functions that can be used to evaluate the trained model in all possible scenarios. The good news is that a lot of them are actually part of scikit-learn's metrics module.

Let's have a quick look at some of the most important scoring functions. ...