It is possible to use the tree model viewer component to distribute an input dataset between the nodes and then clicking on a node will display the data in that node - this is really useful!

However, it would be better if the graphical data of the tree nodes was updated to reflect the data actually in the nodes rather than the data from the training set.

For example in the screenshot below, node 2 originally contained 13 "low" samples" and you can see this when you hover the cursor over the node. However, for the input dataset there is no way of knowing how many of each class are in a particular node without clicking on it, e.g. on clicking node 2 I find there are no samples at all in here.

I think it would be better to display the number of samples for the input data rather than the training data - is it possible to do this somehow? - does anyone have a modified viewer?