In this chapter, we covered a lot of ground. We began by introducing ensembles, some of the most powerful and popular techniques in competitive machine learning contexts. We covered both the theory and code needed to apply ensembles to our machine learning projects, using a combination of expert knowledge and practical examples.
In addition, this chapter also dedicates a section to discussing the unique considerations that arise when you run models for weeks and months at a time. We discussed what data change can mean, how to identify it, and how to think about guarding against it. We gave specific consideration to the question of how to create sets of models running in parallel, which you can switch between based on seasonal change or performance drift in your model set.
During our review of these techniques, we spent significant time with real-world examples with the specific aim of learning more about the creative mindset and broad range of knowledge required of the best data scientists...