In this chapter, we learned about two branches of machine learning—the supervised and unsupervised learning—and practiced building four machine learning models, each with its pros and cons. Each of those models can be used directly to create an estimate or analyzed to understand the most important features or trends. In many instances, the latter is more important and useful than the estimate itself. While these models are not as hot and complex as others (ahem, neural networks), they are widely adopted and used everywhere—in healthcare, military, engineering, city planning, policy analysis, logistics, and operational management—chances are one of them is running in some form on the device you have in your pocket or the computer that's sitting on your desk, right now.

The particular models we trained in this chapter had the default settings and used raw features we collected from Wikipedia.

In the next chapter, we'll learn how to improve the models—by engineering a better set of features, optimizing the hyperparameters, or switching to a more complex model. Using this process as a starting point, we will also learn to run computational experiments, keeping track of both code and data and ensuring reproducible outcomes.