Internet Advertisements — Decision Trees

Practice with decision trees using the Internet Advertisements dataset from the UCI Machine Learning Repository — deciding whether an image on a web page is an advertisement.

The dataset is deliberately awkward: 3,279 observations described by 1,558 explanatory variables, and heavily imbalanced with 459 advertisements against 2,820 non-advertisements. That imbalance is the point — it is what makes the naive approach look good and be wrong.