How do you visualize/present the predicted output of a random forest model with large trees?

Question

I’ve heard that it’s hard to visualize the output of  random forest models with large trees/forest but I’m finding it hard to understand what the use case for the model is, if you can’t visualize the outputs? How do you use the predictions then? Is there a way to visualize this?

Answer 1

The visualization of a Random Forest especially when you have a lot of features could be a challenge. It is similar to visualizing a set of data points when we have many features. One of the ways is just selecting the most important features by feature selection methods such as PCA and visualize a smaller tree such as the following figure with that smaller set of features, like the following which is taken from this article (you can find the code there).Otherwise, you will end of with a very large tree which does not help anyone, especially if you want to report to the manager or customer. But, when you are going to build the model and deploy, it is better to use it if it gives you better scoring regarding the metrics you choose, and explain the stakeholder the tree looks like the following figure, but the most important features we are deciding on is similar to the tree you see in the previous figure.