Chapter 10: Logistic Regression

In this and the next few chapters, we will explore models for classification. These involve targets with two or several class values, such as whether a student will pass a class or not or whether a customer will choose chicken, beef, or tofu at a restaurant with only these three choices. There are several machine learning algorithms for these kinds of classification problems. We will take a look at some of the most popular ones in this chapter.

Logistic regression has been used to build models with binary targets for decades. Traditionally, it has been used to generate estimates of the impact of an independent variable or variables on the odds of a dichotomous outcome. Since our focus is on prediction, rather than the effect of each feature, we will also explore regularization techniques, such as lasso regression. These techniques can improve the accuracy of our classification predictions. We will also examine strategies for predicting a multiclass...