Balancing Interpretability and Predictive Power with Cubist Models in R