Seeing the Forest for the Trees: Random Forests and Predicting Fracking

When you have thousands of variables that you can select from to describe and predict whether active fracking is occurring, how do you choose? Random forests are a learning method that uses an ensemble of decision trees to build a predictive model. We’ll discuss how to make a random forest and how we used one to tell what measures from the microbiome and the sample site best predicted active fracking status. The talk is geared towards undergraduates and assumes little prior background in either statistics or biology. Presented by Kim Roth, Ph.D., Juniata College