When to Use Decision Trees

Probably the biggest advantage of decision trees is how easy it is to interpret a trained model. After running the algorithm on our example problem, we not only end up with a tree that can make predictions about new users, we also get the list of questions used to make those determinations. From this you can see that, for instance, users who find the site through Slashdot never become paid subscribers, but users who find the site through Google and view at least 20 pages are likely to become premium subscribers. This, in turn, might allow you to alter your advertising strategy to target sites that give you the highest quality traffic. We also learn that certain variables, such as the us....