Coming up with a perfect machine learning model is not simple if you do not use a good testing methodology. This seemingly perfect model will fail the moment you deploy it. Testing the model's performance is not an easy task, but it is an essential part of every data science project. Without proper testing, you can't be sure whether your models will work as expected, and you can't choose the best approach to solve the task at hand.
This chapter will explore various approaches for model testing and look at different types of metrics, using mathematical functions that evaluate the quality of predictions. We will also go through a set of methods for testing classifier models.
In this chapter, we will cover the following topics:
- Offline model testing
- Online model testing
