Hallo.

I have a "little" question :-)
In my current research I faced a problem of creating the decision tree with very unusual aspect (as I see from publications noone was interesting in such problem): I need to create decision tree with more than one target attribute.
It means, that final leafs of the tree will have not values "yes" or "no", which describe some parameter, but values for (for example) 2 parameters:
"param1=yes, param2=no". Goal of decision tree will be to find states for both of these parameters.
For example. Lets consider classical example with "play football", that You for sure familiar with.
We may have following situation:

Outlook Temperature Humidity Wind Play_football Go_shopping
sunny hot high weak yes no
overcast mild normal strong no yes
rain hot normal weak no no
sunny mild low strong yes no
etc...

So I want to build decision tree for both parameters - play football and go shopping. This decision tree shall bring me to the decision about both(!) these parameters as soon as possible. During decision would be great to find as soon as possible at least one answer, i.e. after we know outlook and temperature we may know shall we play football or not, but still we don’t know about shall we go shopping or not.
Sure, I can build 2 decision trees for each of target attributes, each of them will be optimal (minimal) for each attribute, but I want to have one tree which will be not optimal for each of attributes, but will be optimal for them both together.

Noone of papers that I found yet describing such cases, there are a lot of works on how to proceed lees of data or too many data, but all are focusing on case when there is only ONE target attribute.
Maybe you have some suggestions on how to deal?