In this chapter we learned about simple nonlinear models for classification and regression called decision trees. Like the parlor game Twenty Questions, decision trees are composed of sequences of questions that examine a test instance. The branches of a decision tree terminate in leaves that specify the predicted value of the response variable. We discussed how to train decision trees using the ID3 algorithm, which recursively splits the training instances into subsets that reduce our uncertainty about the value of the response variable. We also discussed ensemble learning methods, which combine the predictions from a set of models to produce an estimator with better predictive performance. Finally, we used random forests to predict whether or not an image on a web page is a banner advertisement. In the next chapter, we will introduce our first unsupervised learning task: clustering.