Summary

Logistic regression has been a go-to tool for me for many, many years when I have needed to predict a categorical target. It is an efficient algorithm with low bias. Some of its disadvantages, such as high variance and difficulty handling highly correlated predictors, can be addressed with regularization and feature selection. We went over examples of doing that in this chapter. We also examined how to handle imbalanced classes in terms of what such targets mean for modeling and interpretation of results.

In the next chapter, we will look at a very popular alternative to logistic regression for classification – decision trees. We will see that decision trees have many advantages that make them a particularly good option if we need to model complexity, without having to worry as much about how our features are specified as we do with logistic regression.