PDA

View Full Version : Regression Tree in the book DATA MINING



riverculture
01-19-2009, 12:47 AM
Hi All:

In the second edition of the book DATA MINING by Ian and Eibe, there is a regression tree in Figure 3.7 (b), I am wondering what the percentage in a tree node is and how it is calculated.

For example, the following numbers are taken from the leaf that is connected to the tree node of MMAX on the left side of tree. 29.8 is the average value, 37 is the number of instances, but what is 8.18%?
29.8
(37/8.18%)

Thank you very much

Wen

Mark
01-19-2009, 02:52 AM
Hi Wen,

That number is the root mean squared error of the predictions of the model at that node (constant for a regression tree or linear function for a model tree) expressed as a fraction of the global standard deviation of the target.

Cheers,
Mark.

riverculture
01-19-2009, 06:53 PM
Hi Mark, Thank you very much!