The Surprising Usefulness of the Single Decision Tree

When data scientists think about decision trees it is usually in the context of ensembles of hundreds or thousands trees such as in the gradient boosting machine or random forests. In all the excitement about these powerful but complex learning machines most data analysts have forgotten about their extraordinary ancestor, the CART single decision tree. In this talk we explore ways in which the single tree can yield powerful insight into the structure of data that is actually superior to that yielded by any other learning machine. We step back in time to review what made the single decision tree such a revolutionary analytical tool and we present a number of applications in which t he single decision tree is more effective and more appropriate for the problem at hand than later-developed multi-tree methods. Examples are drawn from e-commerce consumer behavior, consumer insurance billing, detecting undesirable differences between two or populations, detecting serious data errors, unsupervised learning, binning categorical predictors, and the interpretation of complex models.

Agenda:

4:50 pm - 5:00 pm Arrival and socializing
5:00 pm - 5:10 pm Opening
5:10 pm - 6:50 pm Dan Steinberg, "The Surprising Usefulness of the Single Decision Tree"
6:50 pm - 7:00 pm Q&A

Speaker: Dan Steinberg, Choice Analytics
Zoom link
Webinar ID: 844-1756-3365