Practical Predictive Analytics

Random forest versus decision trees

Even though random forests can be better predictors than singly partitioned trees, the issue of interpretability comes into play. Random forests do not generate visual decision trees, and computationally random forests can take quite a long time to run as they work to grow and optimize multiple trees from many variables. Some feel that they act as a black box since there are so many ways to optimize and the underlying methods are not readily transparent. However, it is an optimization method and can be very accurate, if extreme accuracy is your goal.