optimizing Decision tree

Thiru

I would like to go through the step by step result of ID3 decision tree
algorithm. would like to see what would be the value of information gain
if we choose particular feature at specific level. What would happen if
we artifically change the most informative feature to some other at root node.
can we play around Id3 algorithm or if web resources to try that for a sample data set. thanks.

Thiru

thank you.

Telcontar120

RapidMiner is not set up very well to do this kind of step-by-step evaluation of a decision tree because there are no manual controls for variable selection and split values at each tree branch. You can do this somewhat manually for the root node by using the Decision Stump operator, but that will only let you evaluate and compare different candidates for the root node. The only other way to do this easily is to compare different trees after selective removing attributes so that only certain subsets of attributes are available for each tree version.