what do the percentages in the leaves of a decision tree represent?

in figure B), there are leaves (gray boxes) with 3 values, for example, the leftmost leaf has 19.3 (28/8.7%) as its values, the 3 values are (19.3, 28, and 8.7%).

19.3 is the average value of the instances that reach that leaf 28 is the number of instances (out of the 209 total instances) what does the 8.7% mean?

thanks

!]1

• What generated these trees? Sep 8 '19 at 4:06
• its a regression tree, i'm not sure of the exact algorithm but i assume splitting attributes on entropy, its from the textbook "data science - practical machine learning tools and techniques" figure 3.4 pg 72 Sep 8 '19 at 13:13
• That book is connected to the Weka software suite, which I don't know much about. I found waikato.github.io/weka-wiki/not_so_faq/j48_numbers , which seems to answer the corresponding question for classification, but it's not immediately obvious to me what metric it would be measuring (in percentages!) for regression. Note that the rightmost leaf in your example lists 359%. Sep 8 '19 at 18:54
• thanks. yeah, not sure what it means for regression either. maybe i'll email the authors, haha Sep 8 '19 at 19:33